These De Doctorat De

THESE DE DOCTORAT DE

L'UNIVERSITE DE RENNES 1 COMUE UNIVERSITE BRETAGNE LOIRE

ECOLE DOCTORALE N° 605 Biologie Santé Spécialité : Cancerology

Par Dimitrios Doultsinos Targeting IRE1 activity in Glioblastoma Multiforme

Thèse présentée et soutenue à Rennes, le 05.04.2019 Unité de recherche : INSERM U1242

Rapporteurs avant soutenance :

Fabienne Foufelle DR1, INSERM UMRS1138, Paris, France. Stephane Rocchi DR2, INSERM U1065, Nice, France.

Composition du Jury :

Président : Nicolas Pallet Professor, Hôpital Européen Georges-Pompidou, Paris, France. Examinateurs : Sophie Janssens Professor, VIB-UGent Center for Inflammation Research, Gent, Belgium. Nicolas Pallet Professor, Hôpital Européen Georges-Pompidou, Paris, France. Dir. de thèse : Dr Eric Chevet DR1, INSERM U1242, Rennes, France.

Invité(s) Leif A. Eriksson Professor, Department of chemistry and molecular biology, University of Göteborg, Göteborg, Sweden. Targeting IRE1 activity in Glioblastoma Multiforme

D. Doultsinos For my mum and dad. You have shown me hard work and perseverance like no one ever will. You have always supported me through an absurd rollercoaster of a life so far whether I deserved it or not. This would not have been possible without you. Summary

The endoplasmic reticulum (ER) is a membranous intracellular organelle and the first compartment of the secretory pathway. As such, the ER contributes to the production and folding of approximately one-third of cellular proteins, and is thus linked to the maintenance of cellular homeostasis and the fine balance between health and disease. The unfolded protein response (UPR) is an integrated, adaptive biochemical process that controls cell homeostasis and maintains normal physiological function. Accumulation of improperly folded proteins in the ER leads to stress, which may push the UPR past beneficial functions such as reduced protein production and increased folding and clearance, to apoptotic signalling. The UPR and one of its major sensors IRE1 are thus contributory to the commencement, maintenance, and exacerbation of a multitude of disease states, including Glioblastoma multiforme (GBM) making it an attractive global target to tackle conditions sorely in need of novel therapeutic intervention. GBM is the commonest primary CNS tumour with an incidence of 3 per 100000. The disease has a dismal prognosis with patients succumbing to the tumour between 15 and 18 months post diagnosis, with a median 5 year survival at less than 6%. In this thesis, in silico, in vitro and in vivo approaches are utilised to assess whether IRE1 is a major pathophysiological mediator and valid pharmacological target in GBM and whether its modulation may provide novel therapeutic options as an adjuvant disease modifying treatment. It is here shown that IRE1 may play a differential role in GBM pathophysiology through angiogenesis and growth as well as adaptation to chemotherapy and maintenance of differentiated GBM cell phenotype through XBP1s and RIDD signalling. XBP1s signalling promotes macrophage infiltration to the tumour, angiogenesis, invasion and maintenance of a differentiated aggressive phenotype, whilst RIDD may attenuate angiogenetic and invasive properties as well control miRNA environment and cell re-differentiation. IRE1 is assessed as a therapeutic target, by generating translational cellular models of GBM carrying IRE1 modulated genetic variants and testing their sensitisation to Temozolomide in the presence of targeted IRE1 kinase inhibitors by establishing a novel drug discovery pipeline and producing six as yet unknown to impact IRE1 activity, inhibitors. This body of work shows that IRE1 is an integral part of GBM pathogenesis and progression and targeting it may prove beneficial in specific subsets of GBM patients. Résumé

Le réticulum endoplasmique (RE) est un organite membranaire intracellulaire et le premier compartiment de la voie de sécrétion. En tant que tel, le RE contribue à la production et au repliement d’environ un tiers des protéines cellulaires et est donc lié au maintien de l’homéostasie cellulaire. L’UPR est un processus biochimique intégré et adaptatif activé en réponse au stress du RE qui contrôle l'homéostasie cellulaire et maintient une fonction physiologique normale. L'accumulation de protéines mal conformées dans le RE entraîne un stress qui peut pousser l'UPR à la signalisation apoptotique. L'UPR et l'un de ses principaux capteurs IRE1 contribuent ainsi au début, au maintien et à l'exacerbation d'une multitude d'états pathologiques, y compris le glioblastome multiforme (GBM), ce qui en fait une cible thérapeutique novatrice. La GBM est la tumeur primitive du système nerveux central la plus fréquente avec une incidence de 3 sur 100 000. Le pronostic est sombre avec des patients qui succombent à la tumeur entre 15 et 18 mois après le diagnostic, avec une survie médiane à 5 ans inférieure à 6%. Dans cette thèse, des approches in silico, in vitro et in vivo sont utilisées pour déterminer si IRE1 est un médiateur physiopathologique majeur et une cible pharmacologique valide dans le GBM et si sa modulation peut fournir de nouvelles options thérapeutiques comme traitement adjuvant modifiant la maladie. Il est montré ici que IRE1 peut jouer un rôle différentiel dans la physiopathologie des GBM par le biais d'angiogenèse et de croissance, ainsi que dans l'adaptation à la chimiothérapie et le maintien du phénotype différencié des cellules GBM par le biais de XBP1 et de la signalisation RIDD. La signalisation XBP1 favorise l'infiltration des macrophages dans la tumeur, l'angiogenèse, l'invasion et le maintien d'un phénotype agressif différencié, tandis que le RIDD peut atténuer les propriétés angiogénétiques et invasives, ainsi que contrôler la ré-différenciation cellulaire et l'environnement. IRE1 est évalué en tant que cible thérapeutique en générant des modèles cellulaires traductionnels de GBM portant des variants génétiques modulés par IRE1 et en testant leur sensibilisation au témozolomide en présence d'inhibiteurs de la kinase IRE1 ciblés, en établissant un nouveau pipeline de découverte de médicaments et en produisant six médicaments non encore impactés. activité, inhibiteurs. L’ensemble de ces travaux montre que IRE1 fait partie intégrante de la pathogenèse et de la progression de la GBM. Son ciblage peut s'avérer bénéfique dans des sous-ensembles spécifiques de patients atteints de GBM. Table of Contents

Preface ...... 8

Introduction ...... 10

Chapter 1: Foreword………………………………………………………………...... …………………………………..13

Chapter 1: Article #1 – “Endoplasmic Reticulum Stress Signalling – from basic mechanisms to clinical applications” (2018) ...... 14

1.1 ER Structure ...... 15

1.2 ER Functions ...... 16

1.3 Perturbing ER functions ...... 17

1.4 ER Stress Consequences ...... 19

1.5 Physiological ER stress signalling ...... 27

1.6 Pharmacological targeting of the UPR ...... 29

1.7 The UPR in the clinic ...... 32

Chapter 1: Conclusions and Contributions ...... 52

Chapter 2: Foreword ...... 54

Chapter 2: Article #2 – “Control of the Unfolded Protein Response in Health & Disease” (2017) ...... 55

2.1 Introduction ...... 55

2.2 The UPR: Canonical and Noncanonical Signalling Pathways ...... 56

2.3 Control of ER Proteostasis in Diseases ...... 58

2.4 The UPR as a Therapeutic Target: From the Tools to the Small Molecules ...... 59

2.5 Molecules Targeting IRE1 and their Use in Disease Models ...... 62

Chapter 2: Conclusions and Contributions ...... 69

Chapter 3: Foreword ...... 71

Chapter 3: Glioblastoma multiforme and IRE1 – clinical implications of translational research ...... 72

3.1 History of GBM ...... 72

1 3.2 Diagnosis, Prognosis and current therapeutic interventions...... 73

3.3 Current clinical trials in GBM ...... 77

3.4 IRE1 involvement in GBM pathophysiology ...... 78

Chapter 3: Conclusions and Contributions ...... 88

Hypothesis and Objectives ...... 89

Results ...... 90

Chapter 4: Foreword..……………………………………………………………...... ………………………………….93

Chapter 4: Article #3 – “Dual IRE1 RNase functions dictate glioblastoma development” (2018)…………………………………………………………………………………..……………………………………………….94

Chapter 4: Conclusions and Contributions…….. ..…………………...... …………………………………141

Chapter 5: Foreword..……………………………………………………………...... …………………….…………..143

Chapter 5: Manuscript #4 – “IRE1 signaling maintains glioblastoma cell differentiation through the XBP1s/miR-148a-mediated repressionof stemness transcription factors” (submitted 2019)…………………………………………………………………...... 144

Chapter 5: Conclusions and Contributions………..…………………...... …………………………………184

Chapter 6: Foreword..……………………………………………………………...... …………………………………186

Chapter 6: Manuscript #5 – “Novel IRE1 catalytic inhibitors chemo-sensitise Glioblastoma cells to Temozolomide” (submitted 2019)...... 187

Chapter 6: Conclusions and Contributions………...…………………...... …………………………………231

Discussion ...... 232

Conclusions ...... 251

Epilogue ...... 249

Figure & Table List

Since this thesis is composed of chapters that have already been published and to avoid ambiguity between figures with the same name across different chapters, this list is composed by introducing a numeric identifier before the figure number. Therefore figure 1.1 for the purpose of this list refers to figure 1 in Chapter 1, figure 3.2 refers to figure 2 in Chapter 3 etc. For figures not in designated chapters (i.e. chapters 1-6) the numeric identifier will be replaced by a letter identifier. Therefore figure 1 of the discussion will be figure D.1, of the introduction I.1 etc.

Figures

Preface Figure: Targeting IRE1 activity in Glioblastoma Multiforme………………………………………....……9 Figure I.1: Introduction to Targeting IRE1 activity in Glioblastoma Multiforme………...... ……………….11 Figure 1.1: ER molecular machines and contact sites with other organelles…………………………………16 Figure 1.2: Signalling the UPR and downstream pathways…………………………………………………………….20 Figure 1.3: UPR disease biomarkers and therapeutic targets…………………………………………………………..33 Figure 2.1: The three UPR sensors………………………………………………………………………………………………56 Figure 3.1: GBM timeline…………………………………………………………………………………………………………..73 Figure 3.2: Anatomical GBM incidence………………………………………………………………………………….……75 Figure 3.3: Studying IRE1 in GBM………………………………………………………………………………………….….79 Figure R.1: Results of targeting IRE1 activity in Glioblastoma Multiforme……………………………………..91 Figure 4.1: IRE1 signalling signatures in Glioblastoma Multiforme……………………………….….……………96 Figure 4.2: Impact of somatic mutations on IRE1 signalling………………………………………………….………98 Figure 4.3: Impact of IRE1 somatic mutations on tumour development……………………………………….100 Figure 4.4: XBP1s and RIDD of mRNA signals in GBM……………………………………………………….…….102 Figure 4.5: RIDD of miRNA (miRIDD) signals in GBM…………………………………………………………..…103 Figure 4.6: Deconvolution of IR1 signalling in GBM tumours…………………………………………………..…105 Figure 4.7: Primary GBM lines exhibit IRE1 signalling properties of the parental tumours……..………106 Figure 4.8: Modulating IRE1 activity in primary GBM lines and phenotypic outcomes…………….…….108 Chapter 4 supplemental figures…………………………………………………………………………………….…128-140 Figure 5.1: IRE1 activity is associated with cancer differentiated states in GBM specimens………..……168 Figure 5.2: Genetic perturbation of IRE1 effect on GBM cell reprogramming………………………………169 Figure 5.3: Pharmacological inhibition of IRE1 effect on GBM cell reprogramming………………………170

3 Figure 5.4: Role of IRE1 in stem to differentiated state reprogramming……………………………..…………171

Figure 5.5: XBP1s involvement in GBM cell reprogramming………………………………………….……………172

Figure 5.6: XBP1s-depednent expression of miR-148a prevents GBM cell reprogramming………………173

Figure 5.7: IRE1-depednent control of GBM stemness and reprogramming in vivo…………………..……174

Chapter 5 supplemental figures……………………………………………………………….………………………177-183

Figure 6.1: Candidate in silico library of IRE1 targeting molecules……….………………………………………212

Figure 6.2: Small IRE1-derived peptides affect IRE1 activity in vitro and in cells……………………………213

Figure 6.3: IRE1-targeting small molecules inhibit IRE1 signalling in GBM cells……………………………214

Figure 6.4: IRE1 inhibitors sensitise GBM cells to TMZ………………………………………………………………215

Chapter 6 supplemental figures……………………………………………………………….………………………216-229

Figure D.1: Incorporation of GLP and GMP in pre-clinical and clinical development of an intra- operative method of IRE1 inhibition administration in GBM………………………………………………..…………….247

Figure D.2: Phase I trial of an IRE1 inhibitor in GBM…………………………………………………………..……249

Tables

Table 1.1: UPR modulators……………………………………………………………………………………………….…....…30

Table 1.2: ER stress associated clinical trials…………………………………………………………...... ………….….…34

Table 2.1: Screening approaches that have led to the discovery of new UPR modulators….……………..60

Table 2.2: UPR modulators………………………………………………………………………………………………………..63

Chapter 4 supplemental tables:……………………………………………………………………………………….119-127

Table 6.1: A detailed representation of in silico and in vitro data for each compound yielded by the docking screen………………………………………………………………………………………………………………………………….208 Table 6.2: A representation of projected ADME property values for non FDA approved compounds Z4, Z6 and IRE1 ribonuclease inhibitor MKC8866………………………………………………………………………………..…209

4 CREB3L3, cAMP responsive element- List of Abbreviations binding protein 3 like 3; CSF, cerebrospinal fluid; CTIMP, clinical trial of investigational 4-PBA, 4-phenylbutyric acid; medicinal product; ADME, absorption distribution metabolism DN, dominant negative; and excretion; DTT, dithiothreitol; ALCAM, activated leukocyte cell adhesion molecule; EGFR, epidermal growth factor receptor;

ALS, amyotrophic lateral sclerosis; eIF2a, eukaryotic translation initiation factor 2a; ATF4, activating transcription factor 4; eIF2B, eukaryotic translation initiation ATF6a, activating transcription factor 6 a; factor 2B; ATF6b, activating transcription factor 6 b; EIS, epithelial to mesenchymal transition ATF6f, cytosolic domain of ATF6; inhibiting sextet;

ATP, adenosine triphosphate; EMA, European medicines agency;

ATRX, Alpha Thalassaemia/Mental EMT, epithelial to mesenchymal transition; retardation syndrome X-linked; EORTC, European organisation for BBB, blood brain barrier; research and treatment of cancer;

BBF2H7, cAMP responsive element- ER, endoplasmic reticulum; binding protein 3 like 2; ERa, oestrogen receptor a; BiP, binding immunoglobulin protein ERAD, ER-associated protein degradation; (gene GRP78); ERN1, endoplasmic reticulum to nucleus bZIP, basic-leucine zipper; signalling 1; CART, chimeric antigen receptor T-cell; ERN2, endoplasmic reticulum to nucleus CCR, cancer cell reprogramming; signalling 2;

CDKN2A, cyclin-dependent kinase ERO-1, ER oxidoreductin 1; inhibitor 2A; FDA, food and drug administration; CE, conformite europeene; GADD34, growth arrest and DNA-damage- CHOP, CAAT/enhancer-binding protein inducible 34; (C/EBP) homologous protein; GBM, glioblastoma multiforme; CNS, central nervous system; GLP, good laboratory practice; CRCL, chaperone-rich cell lysate; GMP, good manufacturing practice; CREB, cAMP response element-binding GRP170, glucose regulated protein 170; protein; GRP78, glucose-regulated protein 78;

5 GRP94, glucose regulated protein 94; NGLY1, N-glycanase;

GSH, glutathione; NOS, not otherwise specified;

IBD, inﬂammatory bowel disease; NPR, NADPH-P450 reductase;

IDH, isocitrate dehydrogenase; OASIS, cAMP responsive element-binding protein 3 like 1; IRE1α, inositol-requiring enzyme 1 α; OLIG2, oligodendrocyte transcription IRE1β, inositol-requiring enzyme 1 β; factor; KEL, kell metallo-endopeptidase; OPLS, optimised potentials for liquid KIRA, kinase inhibiting RNase attenuator; simulations;

KO, knock-out; ORAI1, calcium release-activated calcium channel protein 1; KPS, Karnofsky performance status; PDB, protein data bank; LUMAN, cAMP responsive element- binding protein 3 or CREB3; PDI, protein disulﬁde isomerase;

MAM, mitochondria-associated PDX, patient derived xenograft; membrane; p-eIF2a, phospho-eIF2a; MBTPS1, membrane bound transcription PER1, period circadian regulator 1; factor peptidase, site 1; PERK, protein kinase RNA-like (PKR-like) MBTPS2, membrane bound transcription endoplasmic reticulum kinase; factor peptidase, site 2; PET, positron emission tomography. MD, molecular dynamics; PKR, protein kinase RNA-activated; MDM1/SNX13, mitochondrial distribution and morphology 1/sorting nexin 13; PM, plasma membrane;

MGMT, O6-Methylguanine-DNA- POU3F2, brain specific homeobox/ POU methyltransferase; domain protein 2;

MHRA, medicines health regulation PP1, protein phosphatase type 1; authority; PTEN, phosphatase and tensin homolog; mTOR, mammalian target of rapamycin; QC, quality control; NADP, nicotinamide adenine dinucleotide qPCR, quantitative polymerase chain phosphate; reaction; N-ATF6, N-terminal portion of ATF6 or RB1, RB transcriptional corepressor 1; ATF6f; REIMS, rapid evaporative ionisation mass NCI, national cancer institute; spectrometry; NCIC, national cancer institute of Canada; RER, rough endoplasmic reticulum; NF1, neurofibromin 1; RIDD, regulated IRE1-dependent decay; NF-Y, nuclear transcription factor Y; ROS, reactive oxygen species;

6 SALL2, spalt like transcription factor 2;

SEC22b, vesicle-trafﬁcking protein SEC22b;

SER, smooth endoplasmic reticulum;

SERCA, sarco/endoplasmic reticulum ATPase Ca2+-ATPase;

SOX2, sex determining region Y-box 2;

SPARC, secreted protein acidic and cysteine rich;

SVZ, sub ventricular zone;

TAD, transcriptional activation domain;

TCGA, the cancer genome atlas;

TERT, telomerase reverse transcriptase;

TF, transcription factor;

TMZ, temozolomide;

TP53, tumour protein 53;

TRAF2, tumour necrosis factor receptor- associated factor 2;

TUDCA, tauroursodeoxycholic acid;

UDCA, ursodeoxycholic acid;

UPR, unfolded protein response;

VEGF-A, vascular endothelial growth factor A;

WHO, world health organisation;

WT, wild-type;

XBP1, X-box binding protein 1.

XBP1s, spliced isoform of XBP1;

XBP1u, unspliced isoform of XBP1;

7 Preface

ޡɏɖɣɗɎɘɝɄɐɉɕɢǤȥɉ߿ Ɉ ޡǡݘ Ɉޣɕɉ߿ɖɄɗɛɄɐɉɖ ޡɏɄɎɖާɘݷɓީɘǡݘ Ɉ ޡǡݸ ɈޣəɡɝɒɋɑɄɏɖ ޡݾ ɅɣɔɘɅɖɄɝީɘǡݘ Ɉ“ ɔރ ɑɦɒɔɒ݌ɄɚəާɒɕɄɖɡɝɉɎɒəޟ ɈɡɔɒəɄɕɔɎࠎɒəɄǡܻɐɐޟ ɏɄޥ əާɒɒɔɗɡɔɒəɄǡɏɄޥ əɔީɘɕɄɖɉɦɒəɄɘǡɏɄޥ əޟ ݏɓɟɌɉɒǤ”

ȪɕɕɔɏɖɠəɋɘəɋɘȫɟǡȢɛɔɖɎɗɑɔɣ, ~400 ȱǤȷ.

Almost 2500 years have passed since the father of medicine encapsulated in two sentences the essence of medicine and medical research as we understand it today. In an age when physicians and medical researchers enjoy previously unfathomable riches of resources and expertise, the conundrums medical science is presented with are infinitely more complex. Science never fails to demonstrate that no matter how far our knowledge is advanced, the intricacies of human physiology are of such magnitude that more questions are generated as soon as discoveries are made. As such Hippocrates remains correct: Life for those plagued by disease is short, our quest to aid them is ever going and the time in which to achieve it is infinitesimally short. At the same time though, experience through experimentation is vastly enriched and thus judgement cautiously more relied upon. As experience and judgement afford better healthcare so does opportunity for ideal medical practice arise. This is indeed, as Hippocrates notions, centred upon personalised medicine where professionals from multiple disciplines harmoniously collaborate to provide what is best for the patient, the patient’s supporting network and the general populace.

With this is mind the following thesis is the culmination of three years of multidisciplinary collaboration, where particular mechanisms of protein metabolism were investigated, to provide insight (and thus therapeutically utilisable information) into glioblastoma multiforme (GBM), a fatal central nervous system malignancy. These mechanisms revolve around an evolutionarily conserved transducer of a protein metabolism homeostatic process termed the unfolded protein response (UPR), which resides within the endoplasmic reticulum (ER) of a cell. This transducer is called inositol requiring enzyme 1 (IRE1). Throughout the following manuscript we (this author and you; the reader) will review the status quo of the biology of protein metabolism, the existing mechanisms of drug discovery targeting this biological process and the current understanding of GBM physiology and treatment. Thereafter we will delve into an investigation which generates cellular experimental GBM models of altered IRE1 activity,

8 identifies biological characteristics that link GBM pathophysiology and treatment response to IRE1 biology as well as carry out a structural appraisal of IRE1 itself. These provide a basis to describe how IRE1 may play a role in the development of GBM, how it may play a role in GBM evading chemotherapy and how by targeting IRE1, GBM can be sensitised to current therapies.

The author D. Doultsinos

“Targeting IRE1 activity in Glioblastoma Multiforme”. This is a graphical representation of the evolution of the project described in this manuscript. The role of IRE1 in the pathogenesis of GBM is investigated first by generating cellular experimental GBM models of alteredIRE1 activity, uncovering key roles in cancer cell differentiation, angiogenesis, macrophage infiltration and invasion. At the same time the IRE1 structure was studied to generate therapeutic candidates that can sensitise GBM cells to current chemotherapy treatments.

9 Introduction

One cannot embark upon testing or even forming a hypothesis without acquiring all the pertinent information published by experts in the field. More so in medical research due to the translational value of the investigations and the inherent good practice duties to the potential future applications of such projects in the clinic. Thus extensive literature reviews are necessary to obtain informed views of the field from a number of different perspectives. In this chapter we shall explore just this concept. To fully appreciate the importance of the UPR and subsequently of IRE1 in normal physiology, let alone disease, we shall first go through an in depth account of ER biology (Chapter 1) in the form of a review “Endoplasmic reticulum stress signalling – from basic mechanisms to clinical applications” covering a comprehensive, global view point of ER Stress and its involvement in physiology from development, to growth, to disease and subsequently to the concept of “from bench to bedside and back”. This detailed “analysis of existing ER Stress literature” was co-authored by all the members of the European Union funded Marie Sklodowska Curie Actions (MSCA) Innovative Training Network (ITN) consortium TRAIN-ERS and has served as a checkpoint for a multitude of readerships that want to familiarise themselves with the topic, having already been cited 9 times since its publication on 20/07/18 (Google Scholar). During this first sub-chapter of the introduction we will seek to clarify and construct a landscape of concepts that not only offer a fresh, systematic perspective on the conditions that disrupt ER homeostasis, or ER stress, but also bridge different areas of research in a way that illuminates the importance of ER in the life of an organism. Further testament to the value of collaborative effort, the undertaking of this review not only provided an analysis of the current field of study of ER biology but served as an indispensable exercise of people and time management through the need of coordinating a large number of authors representing a broad geographic and cultural diversity. In Chapter 2 we will next focus a bit further by exploring the relationship between the UPR as is an integrated, adaptive biochemical process and a multitude of disease states, including GBM. We shall look into why the UPR is an attractive global target to tackle conditions sorely in need of novel therapeutic intervention. The accumulation of information on available screen results, readily available therapies, and potential pathways to drug development is the cornerstone of informed clinical research and clinical trial design. In this

10 chapter, published under the title “Control of the Unfolded Protein Response in Health and Disease”, (Published 28/05/17; citations=7 Google Scholar) we will see an in-depth description of the molecules found to target the UPR, through a compilation of all the tools available to screen for and develop novel therapeutic agents that modulate the UPR with the scope of future disease intervention. In the final part of the introduction (Chapter 3) there will be the opportunity to fully inspect GBM as a clinical entity, the current state of the art in its diagnostic and prognostic clinical scope and the potential of therapeutic development to target it. We shall explore the barriers to GBM disease prevention and the reasons for its dismal prognosis all the while attempting to pinpoint the reasons why IRE1 could be a valid therapeutic target in GBM.

“Introduction to Targeting IRE1 activity in Glioblastoma Multiforme”. This is a graphical representation of the logical evolution of the introduction of this manuscript.

11 Chapter 1 1Endoplasmic reticulum stress signalling – from basic mechanisms to clinical applications

1 The artwork for the cover of this issue of The FEBS journal is owned by Wiley.

12 Chapter 1: Foreword

The writing of the following chapter serves a dual purpose. Firstly it provides the opportunity for us to take a global view of the ER as an organelle and thus examine the potential butterfly effect2 of its perturbation on normal physiology and consequently on numerous pathologies. The endoplasmic reticulum (ER) is the site of production and folding of approximately one-third of all cellular proteins; a staggering contribution to cellular and organismal life. Its smooth function is of utmost importance to life in organisms spanning archaea to humans. This functionality oscillates around a resting state which is susceptible to change and hence stress, through a multitude of factors, which could be as benign as a seasonal temperature change or the division of a cell during development. Specific signalling pathways, such as the unfolded protein response (UPR), are responsible for responding to and controlling the magnitude of these oscillations. How well they respond decides the fate of the cell making the ER and the UPR central to the balance between health and disease; life and death. As such a volume where the ER impact on other organelles and cellular behaviour and its potential exploitation to address unmet biomedical needs is of evident value. Secondly it is the examination and result of an exercise in collaborative writing as a deliverable target of a group of early stage scientists covering a variety of ER centric topics across 14 research projects carried out in 8 European research centres as part of an Innovative Training Network “TRAIN-ERS”, funded by the European Commission, through the Marie Sklodowska Curie Actions. As such we will be able to examine the benefits of interdisciplinary collaboration and subsequent results of such an undertaking through the medium of an extensive literature review.

2 Lorenz, E. N. (1972). Predictability: Does the Flap of a Butterfly’s Wings in Brazil Set off a Tornado in Texas. American Association for the Advancement of Science.

13 REVIEW ARTICLE Endoplasmic reticulum stress signalling – from basic mechanisms to clinical applications Aitor Almanza1, Antonio Carlesso2, Chetan Chintha1, Stuart Creedican3, Dimitrios Doultsinos4,5, Brian Leuzzi1, Andreia Luıs6, Nicole McCarthy7, Luigi Montibeller8, Sanket More9, Alexandra Papaioannou4,5, Franziska Puschel€ 10, Maria Livia Sassano9, Josip Skoko11, Patrizia Agostinis9, Jackie de Belleroche8, Leif A. Eriksson2, Simone Fulda7, Adrienne M. Gorman1 , Sandra Healy1, Andrey Kozlov6, Cristina Munoz-Pinedo~ 10 , Markus Rehm11, Eric Chevet4,5 and Afshin Samali1

1 Apoptosis Research Centre, National University of Ireland, Galway, Ireland 2 Department of Chemistry and Molecular Biology, University of Gothenburg, Goteborg,€ Sweden 3 Randox Teoranta, Dungloe, County Donegal, Ireland 4 INSERM U1242, University of Rennes, France 5 Centre de Lutte Contre le Cancer Eugene Marquis, Rennes, France 6 Ludwig Boltzmann Institute for Experimental and Clinical Traumatology, AUVA Research Centre, Vienna, Austria 7 Institute for Experimental Cancer Research in Paediatrics, Goethe-University, Frankfurt, Germany 8 Neurogenetics Group, Division of Brain Sciences, Faculty of Medicine, Imperial College London, UK 9 Department Cellular and Molecular Medicine, Laboratory of Cell Death and Therapy, KU Leuven, Belgium 10 Cell Death Regulation Group, Oncobell Program, Bellvitge Biomedical Research Institute (IDIBELL), Barcelona, Spain 11 Institute of Cell Biology and Immunology, University of Stuttgart, Germany

Keywords The endoplasmic reticulum (ER) is a membranous intracellular organelle endoplasmic reticulum; proteostasis; and the first compartment of the secretory pathway. As such, the ER con- signalling pathway; stress tributes to the production and folding of approximately one-third of cellular proteins, and is thus inextricably linked to the maintenance of cellular Correspondence E. Chevet, INSERM U1242, Centre de Lutte homeostasis and the fine balance between health and disease. Specific ER Contre le Cancer Eugene Marquis, Avenue stress signalling pathways, collectively known as the unfolded protein de la bataille Flandres Dunkerque, 35042 response (UPR), are required for maintaining ER homeostasis. The UPR Rennes, France is triggered when ER protein folding capacity is overwhelmed by cellular

Abbreviations 4-PBA, 4-phenylbutyric acid; ALS, amyotrophic lateral sclerosis; ATF4, activating transcription factor 4; ATF6f, cytosolic domain of ATF6; ATF6a, activating transcription factor 6 a; ATF6b, activating transcription factor 6 b; BBF2H7, cAMP responsive element-binding protein 3 like 2; BiP, binding immunoglobulin protein (gene GRP78); bZIP, basic-leucine zipper; CHOP, CAAT/enhancer-binding protein (C/EBP) homologous protein; CRCL, chaperone-rich cell lysate; CREB3L3, cAMP responsive element-binding protein 3 like 3; CREB, cAMP response element-binding protein; eIF2B, eukaryotic translation initiation factor 2B; eIF2a, eukaryotic translation initiation factor 2a; ERAD, ER- associated protein degradation; ER, endoplasmic reticulum; ERN1, endoplasmic reticulum to nucleus signalling 1; ERN2, endoplasmic reticulum to nucleus signalling 2; ERO-1, ER oxidoreductin 1; ERa, oestrogen receptor a; GADD34, growth arrest and DNA-damage- inducible 34; GRP78, glucose-regulated protein 78; GSH, glutathione; IBD, inflammatory bowel disease; IRE1a, inositol-requiring enzyme 1 a; IRE1b, inositol-requiring enzyme 1 b; LUMAN, cAMP responsive element-binding protein 3 or CREB3; MAM, mitochondria-associated membrane; MBTPS1, membrane bound transcription factor peptidase, site 1; MBTPS2, membrane bound transcription factor peptidase, site 2; MDM1/SNX13, mitochondrial distribution and morphology 1/sorting nexin 13; mTOR, mammalian target of rapamycin; N-ATF6, N-terminal portion of ATF6 or ATF6f; NF-Y, nuclear transcription factor Y; NGLY1, N-glycanase; NPR, NADPH-P450 reductase; OASIS, cAMP responsive element-binding protein 3 like 1; ORAI1, calcium release-activated calcium channel protein 1; PDI, protein disulfide isomerase; p-eIF2a, phospho-eIF2a; PERK, protein kinase RNA-like (PKR-like) endoplasmic reticulum kinase; PKR, protein kinase RNA-activated; PM, plasma membrane; PP1, protein phosphatase type 1; qPCR, quantitative polymerase chain reaction; RER, rough endoplasmic reticulum; RIDD, regulated IRE1-dependent decay; ROS, reactive oxygen species; SEC22b, vesicle-trafficking protein SEC22b; SERCA, sarco/ endoplasmic reticulum ATPase Ca2+-ATPase; SER, smooth endoplasmic reticulum; TAD, transcriptional activation domain; TRAF2, tumour necrosis factor receptor-associated factor 2; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid; UPR, unfolded protein response; WT, wild-type; XBP1s, spliced isoform of XBP1; XBP1u, unspliced isoform of XBP1; XBP1, X-box binding protein 1.

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 241 Federation of European Biochemical Societies. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

14 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

Fax: +33 (0)299253164 demand and the UPR initially aims to restore ER homeostasis and normal Tel: +33 (0)223237258 cellular functions. However, if this fails, then the UPR triggers cell death. E-mail: [email protected] In this review, we provide a UPR signalling-centric view of ER functions, A. Samali, Apoptosis Research Centre, from the ER’s discovery to the latest advancements in the understanding Biomedical Sciences, NUI Galway, Dangan, Galway, Ireland of ER and UPR biology. Our review provides a synthesis of intracellular Fax: +353 91 494596 ER signalling revolving around proteostasis and the UPR, its impact on Tel: +353 91 492440 other organelles and cellular behaviour, its multifaceted and dynamic E-mail: [email protected] response to stress and its role in physiology, before ﬁnally exploring the potential exploitation of this knowledge to tackle unresolved biological Aitor Almanza, Antonio Carlesso, Chetan questions and address unmet biomedical needs. Thus, we provide an inte- Chintha, Stuart Creedican, Dimitrios grated and global view of existing literature on ER signalling pathways Doultsinos, Brian Leuzzi, Andreia Luıs, Nicole McCarthy, Luigi Montibeller, Sanket and their use for therapeutic purposes. More, Alexandra Papaioannou, Franziska Puschel,€ Maria Livia Sassano and Josip Skoko contributed equally to this work and are listed in alphabetical order.

(Received 18 March 2018, revised 24 June 2018, accepted 18 July 2018) doi:10.1111/febs.14608

Introduction whereby ER tubules possess a high membrane curvature compared to the sheets of the nuclear envelope The endoplasmic reticulum (ER) is a cellular orga- and cisternae. The ER occupies an extensive cell-type- nelle that was first visualized in chicken fibroblast-like specific footprint within the cell and is in contact with cells using electron microscopy and was described as many other intracellular organelles. It forms physical a ‘delicate lace-work extending throughout the cyto- contact sites with mitochondria named mitochondria- plasm’ [1]. Its current name was coined almost associated membranes (MAMs), which play a crucial 10 years later by Porter in 1954 [2]. The ER appears role in Ca2+ homeostasis [6]. It also comes in contact as a membranous network of elongated tubules and with the plasma membrane (PM), an interaction regu- flattened discs that span a great area of the cyto- lated by proteins like stromal interaction molecule 1 in plasm [3]. This membrane encloses the ER lumen and the ER and calcium release-activated calcium channel allows for the transfer of molecules to and from the protein 1 in the PM which are controlled by Ca2+ cytoplasm. levels [7]. Vesicle-trafficking protein SEC22b (SEC22b) and vesicle-associated membrane protein 7 are also ER structure involved in the stabilization of ER-PM contacts and PM expansion [8]. The ER also interacts with endo- The ER is classically divided into the rough ER somes [9] and is tethered by StAR-related lipid transfer (RER) and smooth ER (SER), depending on the pres- protein 3 and StAR-related lipid transfer protein 3 ence or absence of ribosomes on the cytosolic face of [10], which also contribute to cholesterol maintenance the membrane respectively. The SER and RER can in endosomes [11]. Interestingly, an ER interaction exist either as interconnected or spatially separated with the endolysosomal system, mediated by the mito- compartments [4]. More recently, a novel classification chondrial distribution and morphology 1/sorting nexin was proposed based on membrane structure rather 13 (MDM1/SNX13) complex [12], suggests ER than appearance. According to this classification, the involvement in autophagy. Indeed, a specialized ER ER comprises the nuclear envelope, sheet-like cisternae structure called the omegasome forms contact sites and a polygonal array of tubules connected by three- with the phagophore, which elongates and becomes a way junctions [5]. A striking difference between these mature autophagosome [13,14] (Fig. 1). In this way, ER structures is the curvature of the membrane, the ER on its own or in coordination with other cell

242 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

15 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

Fig. 1. ER molecular machines and contact sites with other organelles. The ER is primarily subdivided into the SER and RER, with the latter characterized by the presence of ribosomes at its cytosolic surface. Alternatively, the ER has been recently classified into the nuclear envelope, ER sheet-like cisternae and tubular ER (panel 1). The ER forms multiple membrane contact sites with other organelles, including the endosomes and lysosomes (through STARD3, STARD3NL, Mdm1; panel 2), the mitochondria (through Mfn-2, Sig-1R, PERK; panel 3), and the PM (through ORAI1, STIM1, Sec22b, VAMP7; panel 4) with various functional implications. The ER plays instrumental roles in secretory and transmembrane protein folding and quality control, protein and lipid trafficking, lipid metabolism, and Ca2+ homeostasis, all of these processes being mediated by a diverse series of ER resident proteins (schematically depicted in panels 1 and 5). organelles exerts its multifaceted roles in the function- ER [16] where they are exposed to an environment ality of the cell as it is discussed in the next sections. abundant in chaperones and foldases that facilitate their folding, assembly and post-translational modification ER functions before they are exported from the ER [16]. Protein processing within the ER includes signal sequence cleavage, The ER is involved in many different cellular func- N-linked glycosylation, formation, isomerization or tions. It acts as a protein synthesis factory, contributes reduction of disulfide bonds [catalysed by protein disul- to the storage and regulation of calcium, to the synthe- fide isomerases (PDIs), oxidoreductases], isomerization sis and storage of lipids, and to glucose metabolism of proline or lipid conjugation, all of which ultimately [3]. These diverse functions indicate a pivotal role for result in a properly folded conformation [16–19]. Mis- the ER as a dynamic ‘nutrient sensing’ organelle that folded proteins are potentially detrimental to cell func- coordinates energetic fluctuations with metabolic tion and are therefore tightly controlled. Although reprogramming responses, regulating metabolism and protein misfolding takes place continually, it can be cell fate decisions (Fig. 1). exacerbated during adverse intrinsic and environmental conditions. The ER has developed quality control systems to ensure that there are additional opportunities to Protein folding and quality control correct misfolded proteins or, if terminally misfolded, to The ER is involved in secretory and transmembrane be disposed of by the cell. Terminally misfolded secre- protein synthesis, folding, maturation, quality control tory proteins are eliminated by a process called ER- and degradation, and ensures that only properly folded associated degradation (ERAD) [20]. Proteins are first proteins are delivered to their site of action [15]. About recognized by an ER resident luminal and transmem- 30% of all proteins are cotranslationally targeted to the brane protein machinery, then retrotranslocated into

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 243 Federation of European Biochemical Societies.

16 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. the cytosol by a channel named dislocon [21] and the Ca2+ homeostasis cytosolic AAA+ ATPase p97 [22], deglycosylated by N- Ca2+ is involved as a secondary messenger in many glycanase (NGLY1; [23]) and targeted for degradation intracellular and extracellular signalling networks, via the ubiquitin–proteasome pathway [20,24,25] playing an essential role in gene expression, protein (Fig. 1). synthesis and trafficking, cell proliferation, differentiation, metabolism or apoptosis [33]. ER, as the main Lipid synthesis cellular compartment for Ca2+ storage, plays a pivotal role in the regulation of Ca2+ levels and reciprocally The ER also plays essential roles in membrane produc- many ER functions are controlled in a Ca2+-depen- tion, lipid droplet/vesicle formation and fat accumula- dent way, thereby regulating the calcium homeostasis tion for energy storage. Lipid synthesis is localized at of the whole cell [34]. Consequently, both ER and membrane interfaces and organelle contact sites, and cytosolic Ca2+ concentrations need to be highly spa- the lipid droplets/vesicles are exported in a regulated tiotemporally regulated in order for the ER to main- fashion. The ER dynamically changes its membrane tain a much increased physiological intraluminal Ca2+ structure to adapt to the changing cellular lipid con- concentration and oxidizing redox potential than the centrations. The ER contains the sterol regulatory ele- cytoplasm. To modulate these levels, the ER employs ment-binding protein family of cholesterol sensors a number of mechanisms that control Ca2+ concentra- ensuring cholesterol homeostasis [26]. This compart- tion on both sides of the membrane: (a) ER membrane ment also hosts enzymes catalysing the synthesis of cell ATP-dependent Ca2+ pumps for cytosol-to-lumen membrane lipid components, namely sterols, sphin- transport; (b) ER luminal Ca2+-binding chaperones golipids and phospholipids [27]. The synthesis of those for sequestering free Ca2+; and (c) ER membrane lipids from fatty acyl-CoA and diacylglycerols takes channels for the regulated release of Ca2+ into the place at the ER membrane [28], which also hosts 3- cytosol. These mechanisms are facilitated by a tight hydroxy-3-methyl-glutaryl-coenzyme A reductase, the communication between the ER and other organelles, rate-limiting enzyme of the mevalonate/isoprenoid such as the PM and the mitochondria, thereby sup- pathway that produces sterol and isoprenoid precur- porting the cell needs. sors [29]. Precursors made by ER membrane-localized Traditionally thought as a site of protein synthesis, enzymes are subsequently converted into structural recent evidence has established the involvement of the lipids, sterols, steroid hormones, bile acids, dolichols, ER in many different cellular functions: from novel prenyl donors and a myriad of isoprenoid species with roles in lipid metabolism to connections with key functions for cell metabolism. Interestingly, cytoskeletal structures or roles in cytoplasmic stream- MAMs have been identified as a privileged site of sph- ing, our view of the ER keeps rapidly expanding, plac- ingolipid synthesis [30] (Fig. 1). ing it increasingly as a key organelle governing the whole cellular metabolism. ER export Most of the proteins and lipids synthesized in the ER Perturbing ER functions must be transported to other cellular structures, Conditions that disrupt ER homeostasis create a cellu- which occurs mostly through the secretory pathway. lar state commonly referred to as ‘ER stress’. The cel- To maintain the constant anabolic flux, export needs lular response to ER stress involves the activation of to be tightly regulated, and defects in secretion can adaptive mechanisms to overcome stress and restore lead to serious structural and functional consequences ER homeostasis. This response is dependent on the for the ER. Central to this export process is the gen- perturbing agent/condition and the intensity/duration eration of ER COPII transport vesicles, named after of the stress [35]. the family of proteins that shapes and coats them [31]. In addition to COPII vesicle transport, several other mechanisms of lipid export have been Intrinsic ER perturbations described. A variety of lipids can be transported by nonvesicular mechanisms; for example, large lipopro- Cell autonomous mechanisms can lead to ER pertur- tein cargo has been shown to be exported out of the bation and examples of this can be seen in several dis- ER in another type of vesicle termed prechylomicron eases, including cancer, neurodegenerative diseases and transport vesicles [32] or to accumulate in lipid dro- diabetes. The hallmarks of cancer such as genetic plets (Fig. 1). instability and mutations [36] can result in constitutive

244 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

17 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling activation of ER stress response pathways leading to whereas dithiothreitol inhibits protein disulfide bond cell growth, proliferation, differentiation and migra- formation[52]. Alternatively, Brefeldin A impairs ER- tion. In addition, the uncontrolled, rapid growth of to-Golgi trafficking, thus causing a rapid and reversi- cancer cells requires high protein production rates with ble inhibition of protein secretion [53]. Targeting the a consequent impact on ER systems [37]. Many can- Sarco/ER Ca2+-ATPase (SERCA) with compounds, cers have a high mutation load which results in an such as thapsigargin and cyclopiazonic acid [54,55], intrinsically higher level of ER stress. For example, induces ER stress by reducing ER Ca2+ concentration melanoma has the highest mutation burden of any and impairing protein folding capacity. cancer and the sheer numbers of mutated proteins are a source of intrinsically higher ER stress levels. In Exposure to enhancers of ER homeostasis chronic myeloid leukaemia, the fusion protein produced the Philadelphia chromosome, BCR-ABL1, is a Conversely, other molecules have been found that can constitutively active oncoprotein that enhances cell alleviate ER stress. These include small molecules, pep- proliferation and interferes with Ca2+-dependent tides and proteostasis regulators. The frequently used apoptotic response [38]. In addition, mutation-driven 4-phenylbutyric acid (4-PBA) reduces the accumulation ER stress can also induce senescence that contributes of misfolded proteins in the ER [56]. Tauroursodeoxy- to chemoresistance [39]. ER stress has also been linked cholic acid (TUDCA) is an endogenous bile acid able to several neurodegenerative diseases. For example, to resolve ER stress in islet cells [57]. TUDCA is the mutations in the ER resident vesicle-associated mem- taurine conjugate of ursodeoxycholic acid (UDCA), an brane protein-associated protein B in familial amy- FDA-approved drug for primary biliary cirrhosis that otrophic lateral sclerosis (ALS) are linked to induction is also able to alleviate ER stress [58]. The precise of motor neuron death mediated by the alteration of mode of action of such proteostasis modulators still ER stress signalling [40,41]. On the other hand, secre- remains elusive. tory cells such as pancreatic b cells have a highly developed ER to manage insulin production and Temperature release in response to increases in blood glucose. The C96Y insulin variant leads to its impaired biogenesis Body temperature is crucial for the viability of meta- and ER accumulation in the Akita mouse. As the ER zoans; normal mammalian physiological temperatures cannot cope with the mutation induced stress, beta are 36–37 °C. Deviations from this range can disrupt cells die and type 1 diabetes develops [42,43]. Insulin cellular homeostasis causing protein denaturation and/ mutation-related ER stress was also reported in neona- or aggregation [59]. Moreover, an acute increase in tal diabetes [44,45]. temperature, known as heat shock, causes the frag- mentation of both ER and Golgi [59]. Heat preconditioning at mildly elevated temperatures (up to 40 °C) Extrinsic perturbations in mammalian cellular and animal models has been shown to lead to the development of thermotolerance, Microenvironmental stress which is associated with an increase in the expression In tumours, the ER stress observed in rapidly prolifer- of several heat shock proteins and ER stress markers ating cells is compounded by the fact that increased [60,61]. In addition, moderate hypothermia (28 °C) proliferation eventually depletes the microenvironment induces mild ER stress in human pluripotent stem of nutrients and oxygen, causing local microenviron- cells, the activation of which may be sufficient to pro- mental stress and resulting in hypoxia, starvation and tect against severe stress through an effect known as acidosis, all of which cause ER stress and perturb pro- ER hormesis [62,63]. tein, and possibly lipid synthesis [46]. Nutrient deprivation, and particularly glucose starvation, at least in Reactive oxygen species production and other part, promotes ER stress by impairing glycosylation. perturbations Several external agents can induce intracellular reactive Exposure to ER stressors oxygen species (ROS) production, and when ROS pro- Several small molecules that induce ER stress through duction exceeds the antioxidant capacity oxidative a variety of mechanisms have been identified [47,48]. stress negatively affects protein synthesis and ER Stressors such as tunicamycin [49,50], or 2-deoxyglu- homeostasis [64]. ROS, including free radicals, are gen- cose [51] target the N-linked glycosylation of proteins, erated by the UPR-regulated oxidative folding

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 245 Federation of European Biochemical Societies.

18 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. machinery in the ER [65] and in the mitochondria [66]. BiP, IRE1 and PERK homodimerize or oligomerize In this context, increased mitochondrial respiration and trans-autophosphorylate to activate their down- and biogenesis promotes survival during ER stress stream pathways [72]. In contrast, BiP dissociation through a reduction of ROS [67]. The ER provides an from AFT6 reveals an ER export motif [73] which facil- oxidizing environment to facilitate disulfide bond for- itates its translocation to the Golgi apparatus [77]. This mation and this process is believed to contribute to as ‘competition model’ of UPR activation assumes that much as 25% of the overall ROS generated [68,69]. BiP acts as a negative regulator of UPR signalling. The interconnection between the ER and ROS is medi- However, other BiP-dependent or independent models ated by signalling pathways which involve glutathione have been proposed (reviewed in [78]; Fig. 2). (GSH)/glutathione disulfide, NADPH oxidase 4, NADPH-P450 reductase, Ca2+, ER oxidoreductin 1 IRE1 signalling (ERO1) and PDI [70]. The latter, in particular, has been found upregulated in the central nervous system In humans, there are two paralogues of IRE1 (IRE1a of Alzheimer’s disease patients thus highlighting the and b), encoded by endoplasmic reticulum to nucleus relevance of these pathways in neurodegenerative dis- signalling 1 and 2 (ERN1 and ERN2), respectively ease [71]. Overall, from the sections above it is appar- [79–81]. Both human IRE1 isoforms share significant ent that directly or indirectly impaired ER function sequence homology (39%) [20]. IRE1a (referred to contributes to disease development and treatment IRE1 hereafter) is ubiquitously expressed; however, resistance. inositol-requiring enzyme 1 b (IRE1b) expression is restricted mainly to the gastrointestinal tract and the ER stress consequences pulmonary mucosal epithelium [82,83]. Ern1 knockout (KO) in mice is embryonic lethal due to growth retar- In response to ER stress, cells trigger an adaptive sig- dation and defects in liver organogenesis and placen- nalling pathway called the unfolded protein response tal development [84] while Ern2 KO mice develop (UPR), which acts to help cells to cope with the stress colitis of increased severity and shorter latency [82] by attenuating protein synthesis, clearing the unfolded/ but are otherwise histologically indistinguishable from misfolded proteins and increasing the capacity of the the Ern2WT mice. BiP dissociation, caused by accu- ER to fold proteins. mulating unfolded proteins, triggers IRE1 oligomerization and activation of its cytosolic kinase domain. The oligomers position in close proximity, in a face- The UPR to-face orientation, enabling trans-autophosphoryla- The UPR is a cellular stress response originating in the tion. This face-to-face configuration is adopted by ER and is predominantly controlled by three major sen- both human and murine IRE1 [85,86]. Phosphoryla- sors: inositol requiring enzyme 1 (IRE1), protein kinase tion in the activation loop of the kinase domain, RNA-activated (PKR)-like ER kinase (PERK) and specifically at Ser724, Ser726 and Ser729, is not only activating transcription factor 6 (ATF6). The ER lumi- necessary to activate its cytosolic RNase domain [87] nal domains of all three ER stress sensors are normally but is also required to initiate recruitment of tumour bound by the ER resident chaperone, heat shock pro- necrosis factor receptor-associated factor 2 (TRAF2) tein A5 [heat shock protein family A (Hsp70) member and JNK pathway signalling [88]. The IRE1 cytosolic 5, also known as glucose-regulated protein 78 (GRP78) domain, which is highly homologous with RNase L and binding immunoglobulin protein (gene GRP78) [89], induces a selective cleavage of dual stem loops (BiP)], keeping them in an inactive state [72,73]. Accu- within the X-box binding protein 1 (XBP1) mRNA mulating misfolded proteins in the ER lumen engage [79,90,91]. Therefore, IRE1, in a spliceosome indepen- BiP thus releasing the three sensors. A FRET UPR dent-manner, but together with the tRNA ligase induction assay, developed to quantify the association RNA 20,30-cyclic phosphate and 50-OH ligase [92–97], and dissociation of the IRE1 luminal domain from BiP catalyses the splicing of a 26 nucleotide intron from upon ER stress [74], demonstrated that the ER luminal human XBP1 mRNA to produce spliced isoform of co-chaperone ERdj4/DNAJB9 represses IRE1 by pro- XBP1 (XBP1s) [90,91]. XBP1s is a basic leucine zip- moting a complex between BiP and the luminal stress- per (bZIP) transcription factor [98–100] and the sensing domain of IRE1a [75]. Moreover, it has unspliced isoform of XBP1 (XBP1u) is unable to acti- recently been reported that another ER luminal chaper- vate gene expression due to lack of a transactivation one, Hsp47, displaces BiP from the IRE1 UPRosome domain [91]. The N-terminal region of XBP1u con- to promote its oligomerization [76]. Once released from tains a basic region and a leucine zipper domain

246 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

19 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

Fig. 2. Signalling the UPR and downstream pathways. The 3ER stress sensors (PERK, IRE1, ATF6) upon release from BiP, PDIA5, 6 initiate signalling cascades through transcription factor production (ATF4, XBP1s, ATF6f) and associated processes such as RIDD, NFjB activation and ERAD to address the misfolded protein load on the ER. By modulating transcriptional output and translational demand the UPR attempts to re-establish ER protein folding homeostasis and promote cell survival. If ER stress cannot be resolved then mechanisms are triggered to promote cell death. involved in dimerization and DNA binding T (threonine) motif which destabilizes proteins (ubiq- [91,98,100,101]. The XBP1u C-terminal region con- uitin-dependent proteolysis) and contributes to its tains a P (proline), E (glutamic acid), S (serine) and short half-life [98,101–103]. The N-terminal region

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 247 Federation of European Biochemical Societies.

20 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. also contains two other domains: a hydrophobic progressively with the severity of ER stress. However, region that targets XBP1u to the ER membrane and this hypothesis needs further experimental validation. a domain that promotes efficient XBP1 splicing [104– Interestingly, IRE1b was found to selectively induce 106] and cleavage [103] by pausing XBP1 translation. translational repression through the 28S ribosomal IRE1-mediated splicing of XBP1 mRNA results in an RNA cleavage [81] demonstrating that IRE1a and open reading frame-shift inducing the expression of a IRE1b display differential activities [148]. Characteriz- transcriptionally active and BP1s [90,91,101]. XBP1u ing RIDD activity, particularly in vivo, has proven dif- has been reported to negatively regulate XBP1s tran- ficult due to the complex challenge of separating the scriptional activity as well as to promote the recruit- RIDD activity from the XBP1 splicing activity of ment of its own mRNA to the ER membrane IRE1. In addition, basal RIDD can only target specific through the partial translation of its N-terminal mRNA substrates, as full activation and subsequent region [107,108]. XBP1s directs the transcription of a targeting of further transcripts requires strong ER wide range of targets including the expression of stress stimuli (Fig. 2). chaperones, foldases and components of the ERAD pathway, in order to relieve ER stress and restore PERK signalling homeostasis [109,110]. However, XBP1s can also participate in the regulation of numerous metabolic path- PERK was identified in rat pancreatic islets as a ser- ways such as lipid biosynthesis [111–113], glucose ine/threonine kinase and, similar to PKR, heme regu- metabolism [114–118], insulin signalling [117,119,120], lated initiation factor 2 alpha kinase and general redox metabolism [121], DNA repair [122] and it control nonderepressible 2, can phosphorylate eIF2a influences cell fate including cell survival [123], cell [149,150]. PERK is ubiquitously expressed in the body differentiation [124–128] and development [126,129– [149] and has an ER luminal domain as well as a cyto- 131]. Although there is strong evidence pointing to a plasmic kinase domain [150]. BiP detachment from the key role for XBP1 in multiple cellular functions, the ER luminal domain leads to oligomerization [72], exact mechanisms by which XBP1 mediates gene trans-autophosphorylation and activation of PERK transactivation are still elusive. Indeed, in addition to [151]. Active PERK phosphorylates eIF2a on serine 51 the known interaction of the XBP1s transactivation [150]. eIF2a is a subunit of the eIF2 heterotrimer domain with RNA polymerase II, other mechanisms [152,153] which regulates the first step of protein syn- could exist. For example, XBP1 can physically inter- thesis initiation by promoting the binding of the initia- act with many other transcription factors such as AP- tor tRNA to 40S ribosomal subunits [154]. However, 1 transcription factor subunit [132], oestrogen recep- eIF2a phosphorylation by PERK inhibits eukaryotic tor a (ERa) [133], GLI-family zinc finger 1 [134], translation initiation factor 2B (eIF2B) activity and SSX family member 4 [134], forkhead box O1 [114], thereby downregulates protein synthesis [155]. Block- ATF6 [135], cAMP response element-binding protein ing translation during ER stress consequently reduces (CREB)/ATF [135] and hypoxia inducible factor 1 the protein load on the ER folding machinery [156]. alpha subunit [136] (Fig. 2). Remarkably, some transcripts are translated more The RNase activity of IRE1 can also efficiently tar- efficiently during PERK-dependent global repression get other transcripts through a mechanism called regu- of translation initiation. The ubiquitously expressed lated IRE1-dependent decay (RIDD) [137]. Analysis of activating transcription factor 4 (ATF4) [157], whose the in vitro RNase activity of wild-type (WT) vs transcript contains short upstream open reading mutant IRE1 led to the discovery of a broad range of frames (uORFs) [158], is normally inefficiently trans- other IRE1 substrates [138,139] and, interestingly, it lated from the protein-coding AUG [159]. However, was noted that IRE1 can also degrade its own mRNA attenuation of translation from uORFs shifts transla- [140]. RIDD is a conserved mechanism in eukaryotes tion initiation towards the protein coding AUG, [137,141–145] by which IRE1 cleaves transcripts con- resulting in more efficient synthesis of ATF4 [158]. taining the consensus sequence (CUGCAG) accompa- ATF4 can then bind to the C/EBP-ATF site in the nied by a stem-loop structure [142,146]. The cleaved promoter of CAAT/enhancer-binding protein (C/EBP) RNA fragments are subsequently rapidly degraded by homologous protein (CHOP)/GADD153 [160] and cellular exoribonucleases [141,147]. RIDD is required induce its expression [158]. ATF4 and CHOP directly for the maintenance of ER homeostasis by reducing induce genes involved in protein synthesis and the ER client protein load through mRNA degradation UPR [161], but conditions under which ATF4 and [137,141,142]. Recently, it has been proposed that CHOP increase protein synthesis can result in ATP there is basal activity of RIDD [138] which increases depletion, oxidative stress and cell death [162]. eIF2a

248 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

21 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling phosphorylation (p-eIF2a) can also directly enhance transcription factor 2 [161] and XBP1 [71], and various the translation of CHOP [163,164] and other proteins other transcription factors such as serum response fac- involved in the ER stress response, as reviewed in tor [181], components of the nuclear transcription factor [165]. For example, growth arrest and DNA-damage- Y (NF-Y) complex [159,162,163], yin yang 1 [163,164] inducible 34 (GADD34) [166,167] is positively regu- and general transcription factor I [165]. Converging lated by eIF2a phosphorylation [168] and likewise with IRE1 and PERK signalling cascades, ATF6 can transcriptionally induced by ATF4 [169] and CHOP also induce the expression of XBP1 and CHOP to [170]. Interestingly, GADD34 interacts with the cat- enhance UPR signalling [30,166,167]. However, ATF6 alytic subunit of type 1 protein serine/threonine phos- is not the only ER-resident bZIP transcription factor. phatase (PP1) [171], which dephosphorylates eIF2a At least five other tissue-specific bZIPs, named Luman, thereby creating a negative feedback loop that antago- cAMP responsive element-binding protein 3 like 1 nizes p-eIF2a-dependent translation inhibition and (OASIS), cAMP responsive element-binding protein 3 restores protein synthesis [169,170,172]. The transla- like 2 (BBF2H7), CREB3L3 and CREB, reviewed in tional arrest induced by p-eIF2a reduces protein load [183], are involved in ER stress signalling (Fig. 2), high- in ER lumen and conserves nutrients, while ATF4 dri- lighting the regulatory complexity this branch of the ER ven expression of adaptive genes involved in amino stress response is subjected to at the organismal level. acid transport and metabolism, protection from oxidative stress, protein homeostasis and autophagy Noncoding RNAs together help the cell to cope with ER stress [173,174]. However, sustained stress changes the adaptive Noncoding RNAs are connected to the three UPR sen- response to a prodeath response and ultimately, the sors with effects on both physiological and pathological phosphorylation status of eIF2a appears to codeter- conditions [184]. These RNA species mostly include mine the balance between prosurvival or prodeath sig- microRNAs (miRNAs) and also long noncoding RNAs nalling [175,176]. This is accomplished by the above (lncRNAs). This additional level of regulation works in mentioned delayed feedback through which the inter- fact in a bidirectional manner. This means that either play of GADD34, ATF4 and CHOP results in the the UPR sensors themselves or their downstream com- activation of genes involved in cell death, cell-cycle ponents can also modulate their expression levels. A arrest and senescence [177–180] (Fig. 2). certain number of miRNAs have been so far recognized to regulate IRE1, which in turn regulates miRNAs through XBP1s at a transcriptional level and through ATF6 signalling RIDD activity via degradation. One miRNA regulates The transcription factor ATF6, which belongs to an PERK expression, while this in turn regulates miRNAs extensive family of leucine zipper proteins [8], is encoded through its downstream targets. ATF6 is also modu- in humans by two different genes: ATF6A for ATF6a lated by miRNAs, but only one miRNA has been [181] and ATF6B for ATF6b [153]. After its activation found under its direct effect. Upstream of IRE1, PERK in the ER and export to the Golgi, it is cleaved by the and ATF6, the BiP chaperone is also regulated by miR- two Golgi-resident proteases membrane bound tran- NAs but does not control any. In addition to miRNAs, scription factor peptidase, site 1 (MBTPS1) and lncRNAs exhibit a similar role regarding the regulation MBTPS1, releasing a fragment of ~ 400 amino acids of UPR factors and vice versa. Their levels change in corresponding to ATF6 cytosolic N-terminal portion accordance to the cell stress status and depending on (ATF6f). ATF6f comprises a transcriptional activation the pathophysiological context lead to distinct cell domain (TAD), a bZIP domain, a DNA-binding fates. This interconnection between noncoding RNAs domain and nuclear localization signals. In the nucleus, and the UPR may contribute to a more complex net- ATF6f induces UPR gene expression [73,182]. Although work but at the same time reveals the existence of fine- the two ATF6 paralogs share high homology [153], tuning mechanisms governing ER stress responses and ATF6b is a very poor activator of UPR genes due to the their effects in cell homeostasis (described in [184]). absence of eight important amino acids in the TAD domain [157]. Indeed, it rather seems to function as an inhibitor by forming heterodimers with ATF6a [10,158]. Proximal impact of UPR activation Interestingly, ATF6 can modulate gene expression by Transcriptional programmes interacting with other bZIPs, such as CREB [159], cAMP responsive element-binding protein 3 like 3 Each branch of the UPR pathway culminates in tran- (CREB3L3) [160], sterol regulatory element-binding scriptional regulation and, together the UPR’s major

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 249 Federation of European Biochemical Societies.

22 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. transcription factors, ATF6f, XBP1s and ATF4, stim- to limit protein build-up [187,191]. Autophagy is a ulate many adaptive responses to restore ER function pathway involved in the degradation of bulk compo- and maintain cell survival [35]. They regulate genes nents such as cellular macromolecules and organelles. encoding ER chaperones, ERAD factors, amino acid It involves target recognition and selectivity, sequester- transport and metabolism proteins, phospholipid ing targets within autophagosomes, followed by the biosynthesis enzymes, and numerous others [185]. In fusion of the autophagosome with the lysosome, where particular, the IRE1–XBP1 pathway is involved in the targets are then degraded by lysosomal hydrolases induction of ER chaperones and capacity control of [187,192]. The direct link between ER stress and ERAD [186] as well as promoting cytoprotection [187] autophagy has been established in both Saccha- and cleaving miRNAs that regulate the cell death- romyces cerevisiae and mammalian cells, where autop- inducing caspases [188]. ATF6f translocates to the hagy plays a solely cytoprotective role. The PERK nucleus where it activate genes involved in protein (eIF2a) and IRE1 (TRAF2/JNK) branches of the folding, processing, and degradation [185]. ATF4, acti- UPR have been implicated in ER stress-induced vated downstream of PERK and p-eIF2a, increases autophagy in mammalian systems to avoid accumula- the transcription of many genes that promote survival tion of lethal disease-associated protein variants [192]. under ER stress. Some of these prosurvival genes IRE1–JNK signalling activates Beclin 1, a key player include genes that are involved in redox balance, and regulator of autophagy, via the phosphorylation amino acid metabolism, protein folding and autophagy of Bcl-2 and the subsequent dissociation from Beclin [189]. 1. This then leads to the activation of ATG proteins required for the formation of the autophagolysosome [193]. Overall, these mechanisms decrease the build-up Translational programmes of improperly folded proteins in the ER thus allowing Translation is directly impacted by UPR activation adaptive and repair mechanisms to re-establish home- under ER stress conditions, particularly by PERK as ostasis. As the amounts of improperly folded proteins described above. It also affects the expression of sev- decrease, the UPR switches off. However, the molecu- eral miRNAs, which may further contribute to transla- lar details of UPR attenuation still remain to be fur- tion attenuation or protein synthesis [35]. It has been ther elucidated. shown that ER stress can regulate the execution phase Overall, the three mechanisms describe above of apoptosis by causing the transient induction of inhi- decrease the build-up of proteins in the ER which bitor of apoptosis proteins (IAPs). Several papers have allows adaptive and repair mechanisms to re-establish reported that cIAP1, cIAP2 and XIAP are induced by homeostasis. As the amounts of improperly folded ER stress, and that this induction is important for cell proteins decrease, the UPR switches off. However, the survival, as it delays the onset of caspase activation molecular details of UPR attenuation remain to be and apoptosis. PERK induction of cIAPs and the further elucidated. transient activity of PI3K–AKT signalling suggest that PERK not only allows adaptation to ER stress, but it Regulation of MAMs also actively inhibits the ER stress-induced apoptotic programme [190]. Mitochondria-associated membranes (MAMs), which are mainly responsible for Ca2+ homeostasis maintenance as well as lipid transport, mediate the interaction Protein degradation between the ER and mitochondria thereby controlling There are two main protein degradation pathways mitochondrial metabolism and apoptosis [194]. MAMs activated by components of the UPR following ER contain many proteins and transporters which mediate stress: ubiquitin–proteasome-mediated degradation via mitochondrial clustering and fusion, such as the dyna- ERAD and lysosome-mediated protein degradation via min-like GTPase mitofusin-2 (MFN2) [195]. MFN2 autophagy. ERAD is responsible for removing mis- interacts with PERK, serving as an upstream modula- folded proteins from the ER and several genes tor and thereby regulating mitochondrial morphology involved in ERAD are upregulated by ATF6f and and function as well as the induction of apoptosis XBP1s [185]. ERAD involves the retrotranslocation of [196]. Furthermore, the cytosolic domain of PERK misfolded proteins from the ER into the cytosol where serves as an ER-mitochondria tether, thus facilitating they are degraded by the proteasome (see above) [187]. ROS-induced cell death [197].The sigma 1 receptor When accumulation of misfolded proteins overwhelms (Sig-1R) is located in the MAMs and forms a complex ERAD, autophagy is induced as a secondary response with BiP. Recent studies show that S1R stabilizes IRE1

250 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

23 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling at the MAMs upon ER stress, promoting its dimeriza- on ATF4 to induce the expression of cytoprotective tion and conformational change, and prolonging the genes [174]. Another pathway regulating energy meta- activation of the IRE1–XBP1 signalling pathway bolism is the nutrient-sensing mammalian target of through its long-lasting endoribonuclease activity. Fur- rapamycin (mTOR) signalling hub. mTOR is associ- thermore, mitochondria-derived ROS stimulates IRE1 ated with the UPR through crosstalk with regulatory activation at MAMs [198]. Another MAM component pathways (reviewed in [206]), and mTOR inhibitors is Bax-inhibitor-1 (BI-1), regulating mitochondrial such as rapamycin lead to the activation of PERK Ca2+ uptake and apoptosis. BI-1 is a negative regula- signalling, thus favouring cell viability [207]. PERK tor of IRE1-XBP1 signalling and in BI-1 deficient cells can also regulate the PI3K–AKT–mTORC1 axis there is IRE1 hyperactivation and increased levels of through the activation of AKT. Furthermore, it was its downstream targets [199]. Apoptosis activation by observed that mTORC2 plays a role in the inhibition the UPR results in mitochondrial membrane permeabi- of PERK through AKT activation [208]. Altogether lization, with the resulting Ca2+ transfer potentially these data suggest that crosstalk between mTOR and triggering mitochondrial cytochrome c release [200]. the UPR is complex and occurs through multiple Less well understood are the interactions of the mito- pathways. chondria with the ER during sublethal ER stress. The latter results in more ER-mitochondria contacts than Lipid metabolism lethal levels of ER stress, allowing for transfer of Ca2+ and enhancement of ATP production through The UPR can also be activated by deregulated lipid increased mitochondrial metabolism [201] (Fig. 1). metabolism. In this regard, the UPR has been shown These evidences demonstrate the importance of the to be activated in cholesterol-loaded macrophages ER-mitochondria communication in regulating the ER resulting in increased CHOP signalling and apoptosis homeostasis and in coordinating the cellular response [209]. Notably, chronic ER stress leads to insulin to ER stress, thereby restoring cellular homeostatic resistance and diabetes in obesity. This is caused by condition or leading towards cell death. alterations in lipid composition which lead to inhibition of SERCA activity and hence ER stress [210]. On the other hand, the UPR is involved in systemic Redox homeostasis metabolic regulation. Disturbance of ER homeostasis Oxidative stress can be induced through several mech- in the liver is involved in hepatic inflammation, anisms and is critically controlled by the UPR. PERK steatosis and nonalcoholic fatty liver disease [211]. activity helps to maintain redox homeostasis through The PERK–eIF2a pathway has been reported to reg- phosphorylation of NRF2 which functions as a tran- ulate lipogenesis and hepatic steatosis. Compromising scription factor for the antioxidant response [202]. eIF2a phosphorylation in mice by overexpression of ATF4 also regulates redox control and has been GADD34 results in reduced hepatosteatosis upon shown to protect fibroblasts and hepatocytes from high-fat diet [212]. ATF4 the downstream effector of oxidative stress [173], as well as ensuring that there is PERK–eIF2a pathway has also been suggested to an adequate supply of amino acids for protein and regulate lipid metabolism in hepatocytes in response GSH biosynthesis [203]. However, in neurons and to nutritional stimuli by regulating expression of HEK293 cells ATF4 was shown to induce cell death in genes involved in fatty acid and lipid production response to oxidative stress while CHOP was reported [213,214]. Furthermore, it has been demonstrated that to induce ERO1-a, resulting in ER Ca2+ release and the IRE1–XBP1–PDI axis links ER homeostasis with apoptosis in macrophages [204]. Direct interactions of VLDL production which plays an important role in PDIs with ER stress sensors, protein S-nitrosylation dyslipidaemia [215]. In addition, XBP1 is required for and ER Ca2+ efflux that is promoted by ROS con- the normal hepatic fatty acid synthesis and it was tribute to redox homeostasis and by extension to the shown that selective XBP1 deletion in mice resulted balance between prosurvival and prodeath UPR sig- in marked hypocholesterolaemia and hypotriglyceri- nalling [205]. As such, these signalling loops are para- daemia [216]. These studies suggest that ER stress mount to normal cellular function. and the UPR are involved in lipid metabolism. Relieving ER stress ameliorates the disease state associated with lipid metabolism alterations, suggesting Global metabolic impact of the UPR that targeting ER stress might serve as a therapeutic It was recently shown that the UPR and mitochon- strategy for treating diseases associated with lipid drial proteotoxic stress signalling pathways converge accumulation.

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 251 Federation of European Biochemical Societies.

24 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

Glucose metabolism for the cell to regulate metabolism through regulating mTOR signalling, lipid homeostasis as well as insulin It has been suggested that in the liver the PERK– signalling. eIF2a pathway is responsible for disruption of insulin signalling caused by intermittent hypoxia, though IRE1–JNK pathways may still play a role [217]. Adi- Downstream impact of UPR activation ponectin is widely regarded as a marker of functional The activation of UPR leads to the modulation of glucose metabolism and as a suppressor of metabolic many cellular pathways, thereby influencing prosur- dysfunctions. In hypoxic and ER-stressed adipocytes, vival mechanisms as well as processes such as prolifer- reduced adiponectin mRNA levels are observed due ation, differentiation, metabolism and cell death. to negative regulation by CHOP [218,219]. In b-cells, it was shown that IRE1 is involved in insulin biosynthesis after transient high glucose levels. However, UPR-associated cell death chronic exposure to high glucose leads to full UPR Following prolonged activation of the UPR, the cellu- induction and insulin downregulation[220]. IRE1 sig- lar response switches from prosurvival to prodeath. nalling was shown to be involved in insulin resistance Several types of cell death, including apoptosis, necro- and obesity through JNK activation. In hepatocytes, sis/necroptosis and autophagic cell death, can be IRE1-dependent JNK activation leads (a) to insulin induced following ER stress. receptor substrate 1 (IRS1) tyrosine phosphorylation (pY896) decrease and (b) to AKT activation leading to an increase of IRS1 phosphorylation (pS307), con- Apoptosis sequently blocking insulin signalling. A role for XBP1 in the pancreas was demonstrated by the fact Unresolved ER stress can lead to the activation of that b-cell-specific XBP1 mutant mice show hypergly- either the intrinsic (mitochondrial) or extrinsic [death caemia and glucose intolerance due to decreased receptor (DR)] pathways of apoptosis. Both pathways insulin release of b-cells [221]. ER stress-induced acti- trigger activation of caspase proteases that dismantle vation of ATF6 in rat pancreatic beta cells exposed the cell, and all of the three branches of the UPR are to high glucose, impairs insulin gene expression and involved in apoptosis. In the extrinsic pathway, the glucose-stimulated insulin secretion. Interestingly, activation of DRs on the PM leads to the recruitment knocking down expression of orphan nuclear receptor of caspases to the DRs and their proximity-induced short heterodimer partner (SHP) previously reported trans-autoactivation. Intrinsic apoptosis involves the to be involved in beta cell dysfunction by downregu- release of cytochrome c (along with other proapoptotic lating expression of PDX-1 and RIPE3b1/MafA factors) from the mitochondria, which promotes the partly mitigated this effect. However, it remains formation of a cytosolic protein complex to activate a unclear how ATF6 induces expression of SHP and caspase cascade. This release is controlled by pro- and whether ATF6 alone can directly regulate the expres- antiapoptotic members of the BCL-2 protein family. sion of insulin, PDX-1 and RIPE3b1/MafA [222]. It In particular, the BH3-only members of the family has been suggested that physiological impact of ER including PUMA, NOXA and BIM are pivotal com- stress with respect to glucose metabolism depends ponents of ER stress-induced apoptosis [224], and cells upon the availability of glucose. Indeed acute glucose deficient in BH3-only proteins are protected against availability in beta cells leads to concerted efforts of ER stress-induced cell death [190]. ER stress leads to each branch of UPR to supply insulin, while chronic transcriptional upregulation of these proapoptotic glucose stimulation leads to depletion of insulin pro- molecules resulting in cytochrome c release. Both the duction and beta cell mass due to apoptosis. More- IRE1 and PERK arms of the UPR have been linked over, chronic fasting conditions in mice have shown to induction of apoptosis during ER stress. In particu- that XBP1s directly activates the promoter of the lar, CHOP, a transcription factor that is downstream master regulator of starvation response, PPARa of PERK, and a direct target of ATF4, has been demonstrating a further link between the UPR and implicated in the regulation of apoptosis during ER glucose and lipid metabolism [223]. Acquiring further stress. As discussed in section PERK signalling knowledge on link between UPR and metabolic sen- CHOP-induced expression of GADD34 promotes sor mechanisms will significantly expand the possibil- dephosphorylation of p-eIF2a reversing translational ity of gaining beneficial metabolic output. Taken inhibition and allowing transcription of genes includ- together this indicates that the UPR arms are critical ing apoptosis-related genes [172]. CHOP activates

252 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

25 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling transcription of BIM and PUMA, while it represses photodynamic therapy (reviewed in [237]), drives a transcription of certain antiapoptotic BCL-2 family danger signalling module resulting in the surface expo- members such as MCL-1 [225]. In addition, the ATF4/ sure of the ER luminal chaperone calreticulin and the CHOP pathway can increase the expression of other exodus of other danger-associated molecular patterns, proapoptotic genes, such as TRAIL-R1/DR4 and eliciting immunogenic cell death (reviewed in [238]). TRAIL-R2/DR5 which promote extrinsic apoptosis [180]. Apart from CHOP, p53 is also involved in the Necroptosis direct transcriptional upregulation of BH3‑ only proteins during ER stress. However, the link between p53 Necroptosis, a programmed form of cell death, is activation and ER stress is unclear [226]. dependent on the activation of receptor-interacting Although IRE1–XBP1s signalling is mainly prosur- protein kinase 1 (RIPK1), RIPK3 and mixed lineage vival, IRE1 can promote apoptosis. Activated IRE1 kinase domain-like (MLKL) protein and has been can interact directly with TRAF2, leading to the acti- linked to ER stress. In an in vivo mouse model of vation of apoptosis signal-regulating kinase 1 (ASK1) spinal cord injury, there is induction of necroptosis and and its downstream targets c-Jun NH2-terminal kinase ER stress, with localization of MLKL and RIPK3 on (JNK) and p38 MAPK [227,228]. Phosphorylation by the ER in necroptotic microglia/macrophages suggest- JNK has been reported to regulate several BCL-2 fam- ing a link between necroptosis and ER stress in these ily members, including the activation of proapoptotic cells [239]. Necroptosis is frequently activated down- BID and BIM, and inhibition of antiapoptotic BCL-2, stream of TNFR1 when apoptosis is blocked [240]. BCL-XL and MCL-1 [229,230]. In addition, p38 This has been linked to ER stress-induced necroptosis MAPK phosphorylates and activates CHOP, which whereby tunicamycin kills L929 murine fibrosarcoma increases expression of BIM and DR5, thereby pro- cells by caspase-independent, death ligand-independent, moting apoptosis [231,232]. In fact, cell death induc- TNFR1-mediated necroptosis [241]. tion in HeLa cells overexpressing CHOP is dependent on its phosphorylation by p38 MAPK [233]. Interest- Autophagic cell death ingly, it was proposed that ER stress and MAPK signalling act in a positive feed-forward relationship, as Endoplasmic reticulum stress has also been connected ER stress induces MAPK signalling which in turn to autophagic cell death. Autophagy not only pro- increases ER stress [234]. IRE1 signalling may also motes cell survival, but can also mediate nonapoptotic contribute to apoptosis induction through prolonged cell death under experimental conditions when apopto- RIDD activity which degrades the mRNA of protein sis is blocked, or in response to treatments that specifi- folding mediators [142]. cally trigger caspase-independent autophagic cell death Interestingly, recent studies indicate a role for miR- [192]. IRE1a mediated TRAF2 and ASK1 recruitment, NAs in the induction of apoptosis following prolonged and subsequent JNK activation mediates autophagy. ER stress. For example, miRNA29a which is induced JNK-mediated phosphorylation of BCL-2 releases during ER stress via ATF4 results in the downregula- Beclin-1 (while XBP1s also transcriptionally upregu- tion of antiapoptotic Bcl-2 family protein Mcl-1, and lates its expression), which interacts with the ULK1 thus promotes apoptosis [235]. miRNA7 has also been complex to promote vesicle nucleation that leads to linked with ER stress-induced apoptosis, where IRE1 the formation of the autophagosome [242]. Activated reduces miRNA7 levels which results in the stability of PERK can induce autophagy through ATF4 by induc- a membrane-spanning RING finger protein, RNF183. ing vesicle elongation while Ca2+ release from the ER RNF183 has an E3 ligase domain that then causes the lumen through the IP3R can relieve mTOR inhibition ubiquitination and subsequent degradation of the anti- on the ULK1 complex [187]. apoptotic member of the BCL-2 family BCL-XL. Fol- lowing prolonged ER stress, increased expression of UPR-associated morphological changes RNF183 via IRE1 leads to increased apoptosis [236]. In the last decade, it also became clear that ER Endoplasmic reticulum stress causes morphological stress can profoundly modify the immunological con- changes in cellular models. Experiments to date have sequences of apoptotic cell death. Accumulating largely focused on the morphologies associated with in vitro and in vivo evidence have highlighted that the apoptotic and autophagic cell death resulting from activation of the PERK arm of ER stress evoked in UPR activation. UPR-regulated flattening and round- response to selected of anticancer therapies (including ing of cells, indicative of cell death, has been observed anthracyclines, oxaliplatin, radiation and in many model systems, with traditional caspase-

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 253 Federation of European Biochemical Societies.

26 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. dependent apoptosis being responsible [200,243–248]. upregulating protein expression [260]. In ATF6/ These morphological changes can be reversed by phys- murine models subjected to intermittent water depriva- iological and pharmacological ER stress relief tion, similar downstream effects were observed, but [247,249]. Both IRE1 and PERK arms of the UPR signalling pathways were not investigated [261]. ER have been implicated in the observed changes stress-inducing agents palmitate and oxysterol 27- [193,243,244,247,249–251]. As described above, pro- hydroxycholesterol both result in a reduction in leptin grammed cell death and its associated morphological (a long-term mediator of energy balance) expression changes have become a focal and much researched and extracellular concentrations. This has been attribu- outcome of the use of UPR-inducing cytotoxic agents. ted, by using ChIP analysis and siRNA knockdowns, An intensively studied consequence of ER stress is to the fact that the PERK downstream target CHOP the epithelial to mesenchymal transition (EMT) and its negatively regulates C/EBPa, transcriptionally down- role in cancer invasion and metastasis. EMT is an regulating its translation and release [262,263]. UPR essential component of tissue repair following wound- activation has been implicated in the hypothalamic ing, allowing for the migration of new healthy cells and brown adipose tissue response to thyroid hormone into any lesions that have occurred. Morphological triiodothyronine (T3). Elevated T3 levels induce the changes indicative of EMT have been observed in mul- UPR downstream of AMPK in the ventromedial tiple cell models under physiologically relevant stress nucleus of the hypothalamus, resulting in decreased (e.g. hypoxia) and pharmacological induction of ER ceramide levels. JNK1 KO revealed that it acts down- stress [252–255]. The IRE1–XBP1 pathway has been stream of this AMPK-dependent activation, possibly reported to negatively regulate the traditional epithelial as a target of IRE1 but to our knowledge no studies marker E-cadherin, while positively regulating the mes- have yet confirmed this [264]. In response to ER stress enchymal marker N-cadherin in models of colorectal, in hepatocytes, CREBH is exported from the ER and breast and pulmonary fibrosis [254,256,257]. Breast cleaved in the Golgi apparatus. The CREBH cytosolic cancer and pulmonary fibrosis models showed an fragment binds to the promoter region of hepcidin and IRE1–XBP1-dependent regulation of mesenchymal transcriptionally upregulates its production [265]. promoting transcription factor SNAIL that is responsi- These examples of UPR-regulated hormone produc- ble for EMT [254,256]. Human mammary epithelial tion and release give scope for further investigation cells undergo EMT in response to PERK activation, into the longer term, system wide effects of UPR sig- and PERK-mediated phosphorylation of eIF2a is nalling outside of the current focuses on cytotoxicity required for invasion and metastasis [258]. Other ER and acute diseases. stress-regulated pathways have been proposed to act in the EMT in cellular models, including autophagy and Physiological ER stress signalling activation of c-SRC kinase in tubular epithelial cells [259] and the compensatory activation of the NRF-2/ It has been established that ER stress signalling is HO-1 antioxidative stress response pathway in HT-29 important in interorganelle and intercellular interac- and DLD-1 colon cancer cells [252]. Therefore, UPR tions. It therefore comes as no surprise that it forms a signalling pathways appear to induce morphological significant network of interactions upon which normal changes indicative of EMT. These data have generated physiology is based. This is not only the case in interest in the field of cancer research where the phar- humans, but is also conserved throughout species and macological inhibition of UPR components might be has been an important fact in the design of experimen- used to reduce tumour invasiveness and metastasis. tal model organisms to further study ER stress signalling and it role in physiology and disease. Hormone production Embryology and development The tissues and cells of the endocrine system responsible for hormone production and extracellular sig- The UPR as the major conduit of ER stress regulation nalling often have a high protein load, resulting in ER has been extensively studied in developmental biology stress and activation of the UPR. OASIS (CREB3L1) in the majority of organisms commonly used in transla- and ATF6a have been shown to regulate arginine tional research. The use of multiple models has been vasopressin (AVP), a potent vasoconstrictor, in murine important in discerning the variable ER stress signalling and rat models [260,261]. Upon dehydration or salt between species, as demonstrated by the discovery that loading in rat models, cleaved active OASIS is protein quality control in mammals is critically depen- observed binding the AVP promoter region, directly dent on ATF6 while the major player in

254 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

27 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

Caenorhabditis elegans and Drosophila melanogaster is development, making ER stress signalling a key regula- IRE1 [182,266]. Mammalian and other embryos tor in the earliest stages of life in all organisms [277]. implanted in vitro or naturally, undergo a multitude of physical, biochemical and cellular stresses involving epi- Growth and differentiation genetic changes as well as a disproportional increase in protein synthesis load that affect cell differentiation, Many cell types experience a high protein load during proliferation and growth.[267]. In zebrafish, transgenic various stages of differentiation and maturation, result- models have been generated to monitor XBP1 splicing ing in ER stress. In several cases, morphological during development and implantation, showing that changes required for the final function of the cell would maternal XBP1s is active in oocytes, fertilized eggs and not be possible without transient activation of the early stage embryos, presenting a potential model for UPR’s cytoprotective mechanisms. Deletion of PERK study of the impact of water pollutants on embryogene- in murine models results in loss of pancreatic b cell sis [268]. It was recently shown that in medaka fish the architecture but not in cell death, and was accompa- JNK and RIDD pathways are dispensable for growth, nied by an increase in b cell proliferation. This mor- with development solely dependent on the XBP1 arm phological change results in a diabetes mellitus-like of IRE1 signalling, thereby supporting the hypothesis pathology and is not a result of increased cell death as that XBP1 and RIDD may be differentially utilized in previously proposed [278]. Various haematopoietic lin- development and homeostasis [269]. In C. elegans it has eages require the activation of the UPR in order to sur- been postulated that the IRE1-XBP1 axis as well as the vive ER stress resulting from production of PERK pathway are responsible for the maintenance of immunoglobulins and lysosomal compartments in cellular homeostasis during larval development [270]. order to reach maturity [279–281]. One physiological Pronephros formation was shown to be BiP dependent function that is indispensable for survival is the innate in Xenopus embryos, where BiP morpholino knock- immune response, and cell differentiation is at its epi- down not only blocked pronephros formation but also centre. The conversion of B lymphocytes to highly attenuated retinoic acid signalling, impacting markers secretory plasma cells is accompanied by a huge expan- such as the Lim homeobox protein [271]. In early sion of the ER compartment, and genetic alterations to mouse development, it was shown that the BiP pro- induce immunoglobulin production are good examples moter is activated in both the trophoectoderm and of the necessity of ER signalling in normal physiology inner cell mass at embryonic day 3.5 and that absence [123]. This is supported by a study that suggests the of BiP leads to proliferative defects and inner cell mass UPR, and the PERK pathway in particular, govern the apoptosis, suggesting it is necessary for embryonic cell integrity of the haematopoietic stem-cell pool during growth and pluripotent cell survival [272]. Furthermore, stress to prevent loss of function [282]. The ability of mouse studies revealed that ER stress proteins such as skin fibroblasts to produce collagens and matrix metal- BiP, GRP94, calreticulin and PDIA3 were downregu- loproteinases (proteins increased at wound sites), along lated in adult neural tissues compared to embryonic with their ability to differentiate into myofibroblasts, ones, suggesting a pivotal role for ER stress signalling provides another example where physiological ER in the development of neural tissues such as the brain stress may drive morphological cellular transition [283]. and retina [273]. Beyond the nervous system, ER stress Although not yet fully characterized, the RIDD path- signalling impairment has repeatedly shown mouse way has been linked to a multitude of physiological embryonic lethality and, in particular in the hepatocel- processes including lysosomal degradation and xenobi- lular system, multiple studies have demonstrated that otic metabolism through cytochrome P450 regulation IRE1 and XBP1 signalling defects lead to fetal liver [284]. At the same time, substrates of regulated hypoplasia, intrauterine anaemia and early antenatal intramembrane proteolysis such as CREBH are pancreatic dysfunction [274]. The UPR is intrinsically involved in normal physiological processes such as glu- linked to the mouse embryonic morula–blastocyst tran- coneogenesis [284]. Another substrate of regulated sition [275] and this, in combination with evidence that intramembrane proteolysis, OASIS, is involved in mul- there is an immediate postnatal downregulation of BiP, tiple stages of bone homeostasis and development. shows that there is an important role for the UPR both Mice lacking OASIS present with severe osteopenia, in early and late gestation [276]. Taking all this evidence which is compounded by the fact that the gene for type into consideration, it is apparent that the correct inte- 1 collagen is an OASIS target [285]. Moreover, osteo- gration of signals both intracellularly and between the blast OASIS expression is controlled by factors essen- developing oocyte, follicular environment and support- tial to osteogenesis (BMP2), pointing to a PERK- ing cumulus cells is absolutely essential for embryonic eIF2a-ATF4 pathway upregulation during osteoblast

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 255 Federation of European Biochemical Societies.

28 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. differentiation, where ATF4 restores deficiencies of IRE1 RIDD activity is responsible for a reduction in PERK null osteoblasts all the while impacting apopto- the mRNA of proinsulin processing proteins, including sis for bone remodelling [251,286]. Furthermore, a link INS1, PC1 and SYP. These effects can be observed in between osteoblast differentiation and hypoxia has cases of XBP1 deficiency and in cases of extensive been established, with decreased vascularization shown UPR activation, highlighting the divergent effects of in OASIS null mice pointing towards a potential role IRE1 RNase activity [119,221,294]. of ER stress in angiogenesis during bone development [287]. This signalling cascade does not only restrict Amino acid metabolism itself to the normal physiology of bone but also modulates UPR signalling in astrocytes and is responsible The UPR is also described to be involved in amino for the terminal, early to mature, goblet cell differentia- acid metabolism. It was recently described that ATF4 tion in the large intestine [288–290]. mediates increased amino acid uptake upon glutamine deprivation [295]. Furthermore, a low protein diet leads to the upregulation of cytokines mediated by Metabolism IRE1 and RIG1 which results in an anticancer The ER is a site of significant metabolic regulation. immune response in tumours [296]. In summary, these The UPR plays a major role in the regulation of gly- findings show the importance of the various UPR colysis and it was recently shown that IRE1 mediates arms in cell metabolism and energy homeostasis with a metabolic decrease upon glucose shortage in neu- effects not only on the cell itself but also on the whole rons, suggesting an important role for the UPR as an cellular environment. adaptive response mechanism in relation to energy metabolism [291]. Moreover, mTOR signalling adjusts Pharmacological targeting of the UPR global protein synthesis, which is a highly energy con- suming process, and thereby regulates energy metabo- Several small molecules have been reported to modu- lism (reviewed in [292]). late (activate or inhibit) one or more arms of the UPR. Importantly, these molecules have shown promising beneficial effects in diverse human diseases Lipid homeostasis (Table 1). X-ray cocrystal structures are now available The ER is heavily involved in lipid homeostasis. Char- for IRE1 and PERK with several endogenous or acteristically, hepatocytes are enriched in SER, because exogenous ligands. The understanding of how small in addition to protein synthesis, these cells also synthe- molecules bind to the active sites and modulate the size bile acids, cholesterol and phospholipids. XBP1 function of IRE1 and PERK will have a profound ablation in murine liver results in hypolipidaemia due impact on the structure-based drug discovery of novel to feedback activation of IRE1 caused by the lack of UPR modulators. Available X-ray structures, in addi- XBP1. Activated IRE1 induces the degradation of tion to mutagenesis analysis of critical amino acids mRNAs of a cohort of lipid metabolism genes via [297], have revealed a variety of unexpected allosteric RIDD, demonstrating the critical role of IRE1–XBP1 binding sites on IRE1 [297–299]. signalling in lipid metabolism and suggesting that targeting XBP1 may be a viable approach to the treat- Pharmacological modulators of IRE1 ment of dyslipidaemias [113]. It was also reported that in hepatocyte-specific IRE1-null mice, XBP1 is IRE1 signalling information along with CHOP/Gal4- involved in very low-density lipoprotein synthesis and Luc cells and UPRE-Luc engineered cells were used to secretion [215]. Interestingly, ATF6 has also been screen large chemical libraries in high throughput shown to have a role in adipogenesis by inducing adi- screening assays for discovery of pathway-selective pogenic genes and lipid accumulation [293]. modulators of IRE1 [300].

Glucose metabolism IRE1 ATP-binding site The UPR is also involved in regulating glucose meta- IRE1 modulators have been discovered primarily by bolism. Initial murine studies suggested the PERK– traditional drug discovery methods, identifying inhibi- eIF2a arm was responsible for impaired insulin sig- tors speciﬁc to the kinase or RNase domain (Table 1). nalling due to knock out effects on beta cells during The IRE1 kinase modulators were used as tools to development. Further studies have since shown that understand the allosteric relationship between the

256 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

29 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

Table 1. Different modulators that target the UPR-transducer protein pathways. Molecule name, respective molecular target and brief description with the associated reference are provided (ND: not determined).

UPR Arm Name Target Brief description Reference

PERK GSK2656157 PERK Kinase In preclinical stage for multiple myeloma [314,364] and pancreatic cancer Salubrinal GADD34/PP1c Inhibition of eIF2a dephosphorylation [365–367] In ALS, it increases lifespan of mutant superoxide dismutase 1 transgenic mice In Parkinson’s disease, it increases neuronal survival of a-synuclein transgenic mice ISRIB eIF2b Decreased ATF4 expression [322] Guanabenz GADD34/PP1c Inhibitor of eIF2a phosphatase, [368] Sephin1 GADD34 (PP1c) Inhibitor of eIF2a phosphatase [369] IRE1 Salicylaldimines IRE1 RNase IRE1aRNase active-site inhibitor [305] STF-083010 IRE1 RNase IRE1a RNase active-site inhibitor [308] In preclinical stage for multiple myeloma treatment MKC-3946 IRE1 RNase IRE1a RNase active-site inhibitor [307,370] In preclinical stage for multiple myeloma treatment 4l8c IRE1 RNase IRE1a RNase active-site inhibitor [306] In preclinical stage for multiple myeloma treatment APY29 IRE1 Kinase IRE1a kinase active-site inhibitor [303] Sunitinib IRE1 Kinase IRE1a kinase active-site inhibitor [85,304] FDA approved for renal cell carcinoma It acts on multiple kinases KIRA IRE1 Kinase IRE1a kinase active-site inhibitor [371] Toyocamycin IRE1 RNase IRE1a RNase active-site inhibitor [309,372] In preclinical stage for various cancers treatment 3-ethoxy-5,6- IRE1 RNase IRE1a RNase active site inhibitor [305] dibromosalicylaldehyde Apigenin Proteasome Increase of IRE1a nuclease activity in model [373] FIRE peptide IRE1 Kinase Modulation IRE1 oligomerization in vitro, [85] Xbp1 mRNA cleavage in vitro, in cell culture and in vivo (Caenorhabditis elegans) ATF6 Apigenin ATF6 Upregulation of ATF6 expression [373] Baicalein ATF6 Upregulation of ATF6 expression [374] Ceapin ND Inhibitor of ATF6 [323] Kaempferol ATF6 Downregulation of ATF6 expression [375] Melatonin ATF6 Inhibitor of ATF6 [325] Compound 147 ATF6 Activator of ATF6 [376] Compound 263 ATF6 Activator of ATF6 [376] 16F16 PDI Inhibitor of PDI [377]

kinase and RNase domains [301,302]. Kinase inhibi- inhibit the phosphorylation but stabilize the active tors can be broadly classed as (a) ATP-competitive form of the kinase domain. An active kinase confor- inhibitors that inhibit the kinase domain and activate mation is seen in human apo dP-IRE1* (PDB 5HGI), the RNase domain and (b) ATP-competitive inhibitors as a back-to-back dimer. Notably, the DFG motif that inhibit the kinase domain and inactivate RNase (Asp711-Phe712-Gly713) faces into the active site (kinase inhibiting RNase attenuators – KIRAs). Avail- (DFG-in), with helix-aC-in conformation. In contrast, able IRE1 crystal structures reveal a possible mecha- human IRE1 bound to KIRA compound 33 (PDB: nism of RNase activation by conformational changes 4U6R) shows an inactive kinase conformation, with that occur in the kinase domain when transitioning DFG-in and helix-aC-out conformation. The inactive from a monomeric to an active dimeric state. Type I conformation is incompatible with back-to-back dimer IRE1 kinase inhibitors include APY29 [303] and suni- formation due to the displaced helix-aC [301]. Imida- tinib [304], which target the ATP-binding site and zopyrazine-based inhibitors and other KIRAs

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 257 Federation of European Biochemical Societies.

30 Compendium of endoplasmic reticulum stress signaling A. Almanza et al. allosterically inhibit the RNase activity of phosphory- was also the first oral small molecule to prevent neu- lated IRE1 by possibly displacing helix-aC from an rodegeneration in vivo in prion-diseased mice, with active conformation to an inactive conformation [301]. GSK2606414 reducing the levels of p-PERK and p- eIF2a and restoring protein synthesis rates [316]. Despite the promising selectivity profile, pharmacologi- IRE1 RNase-binding site cal inhibition of PERK in mice caused damage to exo- IRE1 RNase inhibitors include salicylaldehydes [305] crine cells and pancreatic beta cells, a similar 4l8C [306], MKC-946 [307], STF-83010 [308], toy- phenotype to that observed in PERK/ mice [317]. ocamycin [309] and hydroxyl-aryl-aldehydes [86]. The Furthermore, GSK2606414 and GSK2656157 were reported cocrystal structures of murine IRE1a with found recently to inhibit RIPK1 at nanomolar concen- salicyaldehyde-based inhibitor show that Lys 907 is trations [318]. To overcome the b-cell toxicity, small involved in Schiff base arrangement (PDB code: 4PL3 molecules modulating the eIF2a pathway without [86]). Lys 907 is a crucial residue present within the directly inhibiting PERK were examined. Integrated hydrophobic pocket of the IRE1 RNase catalytic site stress response inhibitor (ISRIB) is the first small [310]. Quercetin is reported to activate IRE1 through a molecule described to bind and activate guanine site distinct from the nucleotide-binding site (crystal nucleotide exchange factor eIF2B [319,320]. Unlike structure PDB 3LJ0), increasing the population of GSK inhibitors, ISRIB did not show any pancreatic IRE1 dimers in vitro [299]. A recent in silico study toxicity [321]. Interestingly, ISRIB increased learning identified the anthracycline antibiotic doxorubicin as and memory in WT mice [322] (Table 1). an inhibitor of the IRE1-XBP1 axis [311]. Covalent binders are very efficient in the sense that they com- ATF6 modulators pletely block the proteins to which they bind, but this can also have several drawbacks [312]. Noncovalent The identification of small molecules that modulate kinase and allosteric modulators in general inhibit ATF6 has been challenging due to lack of potentially competitively and are thus less efficient, but can at the druggable binding sites and unavailability of the pro- same time be extremely useful in obtaining new tein crystal structure. Recently, Walter and colleagues insights for developing selective and potent modulators identified selective inhibitors of ATF6 signalling, the of IRE1a-XBP1 signalling (Table 1). small molecules Ceapins, using a high throughput cell-based screen [323]. Ceapins do not affect the IRE1 and PERK arms of the UPR. Ceapins are Other IRE1 modulators chemically classed as pyrazole amides and extensive Peptides derived from the kinase domain of human biochemical and cell biology evidence show that they IRE1 promote oligomerization in vitro, enhancing trap ATF6 in the ER and thus prevent its transloca- XBP1 mRNA cleavage activity in vitro and in vivo tion to the Golgi upon stress [324]. Ceapins sensitize [85]. However, although peptide-based modulators cells to ER stress without affecting unstressed cells have limited clinical application [313] (Table 1) peptide and hence have potential to be developed within the mimetics may prove more useful. These are different framework of a therapeutic strategy to induce cell aspects that can be exploited to develop selective IRE1 death in cancer cells. A recent study identified mela- modulators. Despite significant progress in understand- tonin as an ATF6 inhibitor, leading to enhanced liver ing IRE1 signalling and in the development of modu- cancer cell apoptosis through decreased COX-2 lators of IRE1 activity, several questions still remain expression [325]. The activation of ATF6 depends on to be answered to fully control IRE1 activity and sig- a redox process involving PDIs suggesting that PDI nalling outcomes, including how to selectively target inhibitors such as PACMA31 [326], RB-11-ca [327], the XBP1 and RIDD arms of IRE1 signalling. P1 [327] and 16F16 [328] may be able to modulate ATF6 activation. Additionally, the serine protease inhibitor 4-(2-aminoethyl) benzenesulfonyl fluoride is Pharmacological modulators of PERK reported to prevent ER stress-induced cleavage of Through biochemical screening of exclusive library col- ATF6 [329] (Table 1). Albeit the above developments lections and structure-based lead optimization, GSK hold strong promise for the future, very little is discovered PERK inhibitors GSK2606414 and known to date about specific binding sites, which GSK2656157 [314]. These potent PERK inhibitors can together with the lack of a crystal structure and be orally administered [314], reducing tumour growth insufficient templates to enable homology modelling, in mouse xenograft models [314,315]. GSK2606414 rational drug design targeting ATF6 remains a

258 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

31 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling challenge. Availability of an ATF6 crystal structure is patients to test the association of 20 SNPs across the in this sense the key aspect, as this will provide ato- XBP1 gene region, it was found that three SNPs mistic level understanding of interactions and mecha- rs5997391, rs5762795 and rs35873774 are associated nism of action, and enable in silico based rational with disease, thus linking cell-specific ER stress design of ATF6 modulators. changes with the induction of organ-specific inflammation. Quantitative changes in ER stress chaperones in The UPR in the clinic the CSF have been proposed as possible biomarkers to monitor the progression of neurodegenerative diseases In this section, we review recent preclinical and clinical such as ALS [338,339]. Finally, the mesencephalic studies in which UPR components were used as dis- astrocyte-derived neurotrophic factor (MANF) can be ease biomarkers or as therapeutic targets (Fig. 3). As used as a urine biomarker for ER stress-related kidney already described in section Perturbing ER functions diseases [340]. MANF localizes in the ER lumen and molecules have been designed to modulate ER stress is secreted in response to ER stress in several cell by inducing the UPR (Brefeldin A, DTT), inhibiting types. Similarly, angiogenin was identified as an ER SERCA Ca2+ ATPases (thapsigargin) or preventing stress responsive biomarker found in the urine of the generation of glycoproteins, and hence, the induc- patients with kidney damage [341]. Thus, noninvasive tion of ER stress through calcium imbalance or mis- ER stress-related biomarkers can be used to stratify folded protein accumulation. They were touted as disease risk and disease development (Fig. 3). potential antitumour therapies as they could potentially induce tumour cell death through ER stress over- ER stress and UPR-based therapies activation. However, none of these compounds were used in the clinic due to their lack of specificity and Beyond their use as biomarkers, ER stress signalling high toxicity. It has been reported though that a pro- components also represent relevant therapeutic targets. drug analogue of thapsigargin, mipsagargin, did dis- BiP was recently recognized as a universal therapeutic play acceptable tolerability and favourable target for human diseases such as cancer and bacte- pharmacokinetic profiles in patients with solid tumours rial/viral infections [333]. Antibodies targeting BiP [330]. On the other hand, section 6 describes molecules exhibited antitumoural activity and enhanced radiation that inhibit the various arms of the UPR. efficacy in non-small-cell lung cancer and glioblastoma multiforme in mouse xenograft models [342]. It was also shown that short-term systemic treatment with a UPR biomarkers monoclonal antibody against BiP suppressed AKT Changes in UPR and ER stress markers in blood or activation and increased apoptosis in mice with tissue biopsy samples can be indicative of disease state endometrial adenocarcinoma [343]. Moreover, the ER- and could be/are utilized as valuable biomarkers for resident GRP94 is being evaluated as a therapeutic different human pathologies. For instance, BiP has target because of its ability to associate with cellular strong immunological reactivity when released into the peptides irrespective of size or sequence [344]. Preclini- extracellular environment [331], and in 1993, it was the cal studies have linked GRP94 expression to cancer first ER stress protein associated with the pathogenesis progression in multiple myeloma, hepatocellular carci- of osteogenesis imperfecta [332]. Since then, further noma, breast cancer and colon cancer. Finally, this evidence suggests overexpression of BiP in several protein has been identified as a strong modulator of human diseases (reviewed in [333]). The UPR tran- the immune system that could be used in anticancer scription factors can also be seen as potential biomark- immunotherapy [345]. ers of various diseases. ATF4 is upregulated and ER stress-induced transcription factors can also rep- contributes to progression and metastasis in patients resent relevant targets. Thus, XBP1s has been one of with oesophageal squamous cell carcinoma [334]. Simi- the main targets for drug discovery and gene therapy larly, XBP1 overexpression is linked to progressive [346]. Elimination of XBP1 improves hepatosteatosis, clinical stages and degree of tumour malignancy in liver damage and hypercholesterolaemia in animal osteosarcoma [335]. In contrast, IRE1–XBP1 downreg- models. As such direct targeting of IRE1 or XBP1 can ulation can differentiate germinal centre B cell-like be a possible strategy to treat dyslipidaemias [113]. In lymphoma from other diffuse large B-cell lymphoma cancer, toyocamycin was shown to inhibit the constitu- subtypes and contributes to tumour growth [336]. tive activation of XBP1s expression in multiple mye- Moreover, XBP1 is genetically linked to inflammatory loma cells as well as in patient primary samples [309]. bowel disease (IBD) [337]. Using cohorts of IBD Despite being the least studied UPR arm, there are

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 259 Federation of European Biochemical Societies.

32 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

Fig. 3. UPR disease biomarkers and therapeutic targets. Schematic representation of the UPR signalling pathway as defined in Fig. 2 and annotated with the relevance to disease of each component. The colour code indicates the type of disease (cancer: orange; metabolic disease: red; degenerative disease: blue; infectious disease: green; inflammatory disease: pink) and the lines indicate the role as biomarker (continuous line) or therapeutic target (dashed line). instances that ATF6 can be a specific clinical target. that expression of orosomucoid-like 3 (ORMDL3) reg- The activation of ATF6 but not IRE1 or PERK has ulates ATF6 expression and airway remodelling been linked with airway remodelling in a mouse model through ATF6 target genes such as SERCA2b, of asthma [347]. Additionally, these studies showed TGFb1, ADAM8 and MMP9 (Fig. 3, Table 2).

260 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

33 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

ER stress targets are also strong candidates for reported as well in different critical care diseases mod- immunotherapy and vaccines development, a good els, such as sepsis [331,332], liver, heart, brain and kid- example of which is the production of chaperone pro- ney ischaemia [353–359] and haemorrhagic shock tein-based cancer vaccines termed chaperone-rich cell [334,335]. But, the pathophysiological impact of ER lysate (CRCL) [348]. The CRCL are purified from stress activation in these conditions severely lacks tumour tissue or recombinantly produced and applied characterization. Multiple factors such as inflamma- as vaccines against murine and canine cancers or infection, hypoxia present in sepsis and shock can induce tious diseases. Advantages of CRCL vaccines include ER stress but its effects are ambivalent. It has been small quantities and easily obtained starting materials shown that induction of ER stress is cytoprotective [349]. Furthermore, DNA vaccination with gp96-pep- [353,354], and that proteostasis promotors/disruptors tide fusion proteins showed increased resistance such as 4-PBA [336] or TUDCA [337] can be used to against the intracellular bacterial pathogen Listeria improve disease outcome. The increase of CHOP in monocytogenes in a mouse model [350]. To improve renal tissue was reported to inhibit inflammatory the efficacy of gp96 vaccines, gp96 was pooled with response in and provide protection against kidney CpG in combination with anti-B7H1 or anti–inter- injury [336]. Moreover, the activation of PERK seems leukin-10 monoclonal antibodies to treat mice with to facilitate survival of lipopolysaccharide-treated car- large tumours [351]. The heterogeneous or allogeneic diomyocytes by promoting autophagy [338]. Addition- gp96 vaccines protected mice from tumour challenge ally, the activation of ATF6 before ischaemia reduced and re-challenge. In addition to its role as a molecular myocardial tissue damage during ischaemia/reperfusion chaperone, GRP94 was likewise identified as a peptide (I/R) injury [339]. Furthermore, induction of BiP in carrier for T-cell immunization [352]. However, the cardiomyocytes stimulated AKT signalling and pro- immunological application of GRP94 derived from its tected against oxidative stress, conferring cellular I/R peptide binding capacity was not further investigated damage protection [340]. In contrast, inhibition of ER (Fig. 3, Table 2). The activation of ER stress has been stress was indicated to limit cellular damage in

Table 2. ER stress-centred clinical trials. A range of clinical entities in endocrinology, oncology and paediatrics have been targeted through clinical trials. This table presents such trials detailing the trial targeted, interventional agent investigated and national authority carrying out the investigation.

Trial Disease Intervention Country

Role of ER stress in the pathophysiology • Diabetes mellitus, type 2 No intervention France of type 2 diabetes • Polycythemia vera ER stress and resistance to treatments • Biological: RNA sample France in Ph-negative myeloproliferative neoplasms Essential thrombocythemia of total leucocytes before start of treatment • Insulin resistance Effect of ER stress on metabolic function • Drug: TUDCA United States Diabetes Other: placebo • Obesity Drug: sodium phenylbutyrate ER stress in chronic respiratory diseases • Chronic airway disorders Observational South Korea • Lung cancer

• HIV-related insulin resistance TUDCA for protease-inhibitor associated • Drug: TUDCA United States insulin resistance Protease inhibitor-related Insulin Other: placebo tablet resistance

ER stress in NAFLD • Obesity Drug: methyl-D9-choline United States • NAFLD

TUDCA in new-onset type 1 diabetes • Type 1 diabetes Drug: TUDCA United States Drug: Sugar Pill (placebo) Effects of Liraglutide on ER stress in • Type 2 diabetes Drug: liraglutide United States obese patients with type 2 diabetes • Wolfram syndrome A clinical trial of dantrolene sodium in • Drug: dantrolene sodium United States paediatric and adult patients with Diabetes mellitus • optic nerve atrophy wolfram syndrome • Ataxia

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 261 Federation of European Biochemical Societies.

34 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

mechanisms for fine-tuning ER signalling, as well as Box 1. First-in-human trial motivated the need for their better characterization Intravenous infusion of GRP78/BiP is safe in patients towards relevant health-related applications. This drive In 2006, Brownlie et al. [362] reported that the prophy- to further ER knowledge has also led to the identifica- lactic or therapeutic parenteral delivery of BiP amelio- tion of emerging roles for the ER in physiology and rates clinical and histological signs of inflammatory disease. In particular, it appears an indispensable tool arthritis in mice. Ten years later, the first human clinical for cellular communication that reaches beyond the trial using intravenous BiP demonstrated that GRP78/ intracellular space. The concept of transmissible ER BiP is safe in patients with active rheumatoid arthritis stress illustrates the far-reaching control that ER sig- and some patients had clinical and biological improve- nalling exerts in interorgan communication affecting ments [363]. In phase I/IIA RAGULA trial, 42 patients disease pathogenesis and normal physiology [360,361]. were screened, and 24 were randomized to receive either Our increasing knowledge of ER signalling mecha- BiP or placebo. The study showed that after a single nisms presents opportunities to exploit the resulting intravenous infusion, BiP may induce remission lasting applications on multiple fronts, including bioengineer- up to 3 months in rheumatoid arthritis patients. ing and health, concepts that may routinely overlap. For example, boosting ER protein production capacity may be applied to cell engineering to increase biologic therapy production. This will drive down costs of biologics, helping demand to be met and leading to more pathologies such as hepatic I/R [341]. This contradic- widely available medications, thus having a significant tion may be due to interference between UPR and effect on public health. Population-wide consequences inflammatory pathways. CHOP-/- mice were reported of ER modulation may not be restricted to the pro- to have more prominent increase in NF-kB activation duction of biologic therapies as its applications could and further upregulation of proinflammatory genes also contribute to bioengineering approaches for crop (CXCL-1, MIP-2, IL-6) [342]. Interestingly, inhibition or livestock improvement. of IRE1-NF-kB by resveratrol protected against sep- A thorough understanding of the ER stress response sis-induced kidney failure [343]. In this light, the mod- and its role in physiology and pathophysiology can be ulation of specific UPR branches is promising applied to develop new ER stress targeted therapies approach for therapy of critical care diseases. and stratifying patients into cohorts suitable for ER- As discussed above, understanding and characteriz- targeted therapies. Considering the enormity of attri- ing the UPR has provided several potential targets to tion rates of novel therapeutic discovery in an ever- develop new therapeutics for various diseases, with an tightening financial climate, there is an urgent need for encouraging increase in the number of clinical trials new therapeutic targets as well as precision tools that based on ER stress pathway targets or associated target and guide innovation to specific patient pools. drugs. Several of these trials [ClinicalTrials.gov, Euro- ER stress signalling may provide such tools. Not only pean Clinical Trials Database and the ISRCTN reg- is it central to life itself but it is involved in a wide istry] have focused on diabetes mellitus. A trial testing array of clinical presentations. Moreover, its effect on TUDCA and 4-PBA for the treatment of high lipid heterogeneous presentations within the same diseases levels or insulin resistance was conducted by the makes it an attractive target for translational precision Washington University School of Medicine; however, medicine. Of course, when undertaking medical although this study was completed in 2014, the find- research or trying to solve a biomolecular functional ings are not available yet. The results from the first mystery one cannot look past the logistical aspect of completed human trial using BiP for rheumatoid the task ahead. The conserved metazoan nature of ER arthritis are described in Box 1. We can anticipate that stress signalling combined with the emergence of high clinical trials to test ER stress targeting drugs in sev- throughput and in silico strategies supplies researchers eral other diseases will shortly ensue. with a wealth of tools to study pathophysiology, from structure to function in multiple in vivo and in vitro Concluding remarks models, producing robust results to be put forward for clinical scrutiny while all the while observing both the The ER has evolved in our knowledge from a key safeguards of the declaration of Helsinki and ethics on player in proteostasis and the secretory pathway to a animal experimentation. Our deeper understanding of cornerstone of metabolic functions. Such wealth of the ER and its major homeostatic regulator, the UPR information has allowed the identification of numerous response, is introducing an individualized molecular

262 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

35 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling approach to health management at a preventative, endolysosomal interorganelle tethering protein. J Cell diagnostic and therapeutic level and, uncovering the Biol 210, 541–551. genetic architecture underlying the ER stress response 13 Hayashi-Nishino M, Fujita N, Noda T, Yamaguchi A, could significantly influence future therapeutic strate- Yoshimori T & Yamamoto A (2010) Electron gies in patients. tomography reveals the endoplasmic reticulum as a membrane source for autophagosome formation. Autophagy 6, 301–303. Acknowledgements 14 Uemura T, Yamamoto M, Kametaka A, Sou Y, We apologize to colleagues whose work was not cited Yabashi A, Yamada A, Annoh H, Kametaka S, due to space limitation. This work was funded by Komatsu M & Waguri S (2014) A cluster of thin tubular structures mediates transformation of the grants from EU H2020 MSCA ITN-675448 (TRAI- endoplasmic reticulum to autophagic isolation NERS) and MSCA RISE-734749 (INSPIRED). CMP membrane. Mol Cell Biol 34, 1695–1706. lab is funded by CERCA, FEDER (A Way To 15 Stefan CJ, Manford AG, Baird D, Yamada-Hanff J, Achieve Europe) and MINECO (BFU2016-78154-R). Mao Y & Emr SD (2011) Osh proteins regulate phosphoinositide metabolism at ER-plasma membrane References contact sites. Cell 144, 389–401. 16 Braakman I & Bulleid NJ (2011) Protein folding and 1 Porter KR, Claude A & Fullam EF (1945) A study of modification in the mammalian endoplasmic reticulum. tissue culture cells by electron microscopy: methods Annu Rev Biochem 80,71–99. and preliminary observations. J Exp Med 81, 17 Hebert DN & Molinari M (2007) In and out of the – 233 246. ER: protein folding, quality control, degradation, 2 Palade GE & Porter KR (1954) Studies on the and related human diseases. Physiol Rev 87, 1377– endoplasmic reticulum. I. Its identification in cells 1408. – in situ. J Exp Med 100, 641 656. 18 Wallis AK & Freedman RB (2013) Assisting oxidative 3 Alberts B, Johnson A, Lewis J, Raff M, Roberts K & protein folding: how do protein disulphide-isomerases Walter P (2002) Molecular Biology of the Cell, 4th couple conformational and chemical processes in edn. Garland Science, New York, NY. protein folding? Top Curr Chem 328,1–34. 4 Borgese N, Francolini M & Snapp E (2006) 19 Aebi M, Bernasconi R, Clerc S & Molinari M (2010) Endoplasmic reticulum architecture: structures in flux. N-glycan structures: recognition and processing in the – Curr Opin Cell Biol 18, 358 364. ER. Trends Biochem Sci 35,74–82. 5 Shibata Y, Voeltz GK & Rapoport TA (2006) Rough 20 Meusser B, Hirsch C, Jarosch E & Sommer T (2005) – sheets and smooth tubules. Cell 126, 435 439. ERAD: the long road to destruction. Nat Cell Biol 7, 6 Hayashi T, Rizzuto R, Hajnoczky G & Su TP (2009) 766–772. MAM: more than just a housekeeper. Trends Cell Biol 21 Hebert DN, Bernasconi R & Molinari M (2010) – 19,81 88. ERAD substrates: which way out? Semin Cell Dev Biol 7 Toulmay A & Prinz WA (2011) Lipid transfer and 21, 526–532. signaling at organelle contact sites: the tip of the 22 Kobayashi T, Tanaka K, Inoue K & Kakizuka A – iceberg. Curr Opin Cell Biol 23, 458 463. (2002) Functional ATPase activity of p97/valosin- 8 Daste F, Galli T & Tareste D (2015) Structure and containing protein (VCP) is required for the quality – function of longin SNAREs. J Cell Sci 128, 4263 control of endoplasmic reticulum in neuronally 4272. differentiated mammalian PC12 cells. J Biol Chem 277, 9 Rowland AA, Chitwood PJ, Phillips MJ & Voeltz GK 47358–47365. (2014) ER contact sites define the position and timing 23 Enns GM, Shashi V, Bainbridge M, Gambello MJ, – of endosome fission. Cell 159, 1027 1041. Zahir FR, Bast T, Crimian R, Schoch K, Platt J, Cox ! 10 Wilhelm LP, Tomasetto C & Alpy F (2016) Touche R et al. (2014) Mutations in NGLY1 cause an STARD3 and STARD3NL tether the ER to inherited disorder of the endoplasmic reticulum- – endosomes. Biochem Soc Trans 44, 493 498. associated degradation (ERAD) pathway. Genet Med 11 Wilhelm LP, Wendling C, Vedie B, Kobayashi T, 16, 751–758. Chenard M, Tomasetto C, Drin G & Alpy F (2017) 24 Kamhi-Nesher S, Shenkman M, Tolchinsky S, Fromm STARD3 mediates endoplasmic reticulum-to- SV, Ehrlich R & Lederkremer GZ (2001) A novel endosome cholesterol transport at membrane contact quality control compartment derived from the – sites. EMBO J 36, 1412 1433. endoplasmic reticulum. Mol Biol Cell 12, 1711–1723. 12 Henne WM, Zhu L, Balogi Z, Stefan C, Pleiss JA & 25 Huyer G, Longsworth GL, Mason DL, Mallampalli – Emr SD (2015) Mdm1/Snx13 is a novel ER MP, McCaffery JM, Wright RL & Michaelis S (2004)

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 263 Federation of European Biochemical Societies.

36 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

A striking quality control subcompartment in endoplasmic reticulum stress signaling in health and Saccharomyces cerevisiae: the endoplasmic reticulum- disease. Am J Physiol Cell Physiol 308, C415–C425. associated compartment. Mol Biol Cell 15, 908–921. 40 Chen H-J, Anagnostou G, Chai A, Withers J, Morris 26 Brown MS & Goldstein JL (1999) A proteolytic A, Adhikaree J, Pennetta G & de Belleroche JS (2010) pathway that controls the cholesterol content of Characterization of the properties of a novel mutation membranes, cells, and blood. Proc Natl Acad Sci USA in VAPB in familial amyotrophic lateral sclerosis. J 96, 11041–11048. Biol Chem 285, 40266–40281. 27 Fagone P & Jackowski S (2009) Membrane 41 Nishimura AL, Mitne-Neto M, Silva HCA, Richieri- phospholipid synthesis and endoplasmic reticulum Costa A, Middleton S, Cascio D, Kok F, Oliveira JRM, function. J Lipid Res 50, S311–S316. Gillingwater T, Webb J et al. (2004) A Mutation in the 28 Yen C-LE, Stone SJ, Koliwad S, Harris C & Farese vesicle-trafficking protein VAPB causes late-onset spinal RV (2008) Thematic review series: glycerolipids. muscular atrophy and amyotrophic lateral sclerosis. Am DGAT enzymes and triacylglycerol biosynthesis. J J Hum Genet 75, 822–831. Lipid Res 49, 2283–2301. 42 Harding HP & Ron D (2002) Endoplasmic reticulum 29 Jo Y & DeBose-Boyd RA (2010) Control of stress and the development of diabetes: a review. cholesterol synthesis through regulated ER-associated Diabetes 51, S455–S461. degradation of HMG CoA reductase. Crit Rev 43 Oyadomari S, Koizumi A, Takeda K, Gotoh T, Biochem Mol Biol 45, 185–198. Akira S, Araki E & Mori M (2002) Targeted 30 Patwardhan GA, Beverly LJ & Siskind LJ (2016) disruption of the Chop gene delays endoplasmic Sphingolipids and mitochondrial apoptosis. J Bioenerg reticulum stress-mediated diabetes. J Clin Invest 109, Biomembr 48, 153–168. 525–532. 31 Barlowe C, Orci L, Yeung T, Hosobuchi M, 44 Colombo C, Porzio O, Liu M, Massa O, Vasta M, Hamamoto S, Salama N, Rexach MF, Ravazzola M, Salardi S, Beccaria L, Monciotti C, Toni S, Pedersen Amherdt M & Schekman R (1994) COPII: a O et al. (2008) Seven mutations in the human insulin membrane coat formed by Sec proteins that drive gene linked to permanent neonatal/infancy-onset vesicle budding from the endoplasmic reticulum. Cell diabetes mellitus. J Clin Invest118, 2148–2156. 77, 895–907. 45 Støy J, Edghill EL, Flanagan SE, Ye H, Paz VP, 32 Siddiqi S, Saleem U, Abumrad NA, Davidson NO, Pluzhnikov A, Below JE, Hayes MG, Cox NJ, Storch J, Siddiqi SA & Mansbach CM (2010) A novel Lipkind GM et al. (2007) Insulin gene mutations as a multiprotein complex is required to generate the cause of permanent neonatal diabetes. Proc Natl Acad prechylomicron transport vesicle from intestinal ER. J Sci USA 104, 15040–15044. Lipid Res 51, 1918–1928. 46 Giampietri C, Petrungaro S, Conti S, Facchiano A, 33 Clapham DE (2007) Calcium signaling. Cell 131, Filippini A & Ziparo E (2015) Cancer 1047–1058. microenvironment and endoplasmic reticulum stress 34 Meldolesi J & Pozzan T (1998) The endoplasmic response. Mediators Inflamm 2015, 417281. reticulum Ca2 + store: a view from the lumen. Trends 47 Jin ML, Park SY, Kim YH, Oh J-I, Lee SJ & Park G Biochem Sci 23,10–14. (2014) The neuroprotective effects of cordycepin 35 Hetz C (2012) The unfolded protein response: inhibit glutamate-induced oxidative and ER stress- controlling cell fate decisions under ER stress and associated apoptosis in hippocampal HT22 cells. beyond. Nat Rev Mol Cell Biol 13,89–102. Neurotoxicology 41, 102–111. 36 Hanahan D & Weinberg RA (2011) Hallmarks of 48 Stechmann B, Bai S-K, Gobbo E, Lopez R, Merer G, cancer: the next generation. Cell 144, 646–674. Pinchard S, Panigai L, Tenza D, Raposo G, Beaumelle 37 Holderfield M, Deuker MM, McCormick F & B et al. (2010) Inhibition of retrograde transport McMahon M (2014) Targeting RAF kinases for protects mice from lethal ricin challenge. Cell 141, cancer therapy: BRAF mutated melanoma and 231–242. beyond. Nat Rev Cancer 14, 455–467. 49 Olden K, Pratt RM, Jaworski C & Yamada KM 38 Piwocka K, Vejda S, Cotter TG, O’Sullivan GC & (1979) Evidence for role of glycoprotein carbohydrates McKenna SL (2006) Bcr-Abl reduces endoplasmic in membrane transport: specific inhibition by reticulum releasable calcium levels by a Bcl-2- tunicamycin. Proc Natl Acad Sci USA 76, 791–795. independent mechanism and inhibits 50 Schultz AM & Oroszlan S (1979) Tunicamycin inhibits calcium-dependent apoptotic signaling. Blood 107, glycosylation of precursor polyprotein encoded by env 4003–4010. gene of Rauscher murine leukemia virus. Biochem 39 Pluquet O, Pourtier A & Abbadie C (2015) The Biophys Res Commun 86, 1206–1213. unfolded protein response and cellular senescence. A 51 Datema R & Schwarz RT (1979) Interference with review in the theme: cellular mechanisms of glycosylation of glycoproteins. Inhibition of formation

264 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

37 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

of lipid-linked oligosaccharides in vivo. Biochem J 184, 65 Bhandary B, Marahatta A, Kim H-R & Chae H-J 113–123. (2012) An involvement of oxidative stress in 52 Cleland WW (1964) Dithiothreitol, a new protective endoplasmic reticulum stress and its associated reagent for SH groups. Biochemistry 3, 480–482. diseases. Int J Mol Sci 14, 434–456. 53 Liu ES, Ou JH & Lee AS (1992) Brefeldin A as a 66 Cadenas E & Davies KJ (2000) Mitochondrial free regulator of grp78 gene expression in mammalian cells. radical generation, oxidative stress, and aging. Free J Biol Chem 267, 7128–7133. Radic Biol Med 29, 222–230. 54 Thastrup O, Cullen PJ, Drøbak BK, Hanley MR & 67 Knupp J, Arvan P & Chang A (2018) Increased Dawson AP (1990) Thapsigargin, a tumor promoter, mitochondrial respiration promotes survival from discharges intracellular Ca2 + stores by specific endoplasmic reticulum stress. Cell Death Differ. inhibition of the endoplasmic reticulum Ca2(+)- https://doi.org/10.1038/s41418-018-0133-4 ATPase. Proc Natl Acad Sci USA 87, 2466–2470. 68 Tu BP & Weissman JS (2004) Oxidative protein 55 Pirot P, Naamane N, Libert F, Magnusson NE, folding in eukaryotes: mechanisms and consequences. Ørntoft TF, Cardozo AK & Eizirik DL (2007) Global J Cell Biol 164, 341–346. profiling of genes modified by endoplasmic reticulum 69 Malhotra JD & Kaufman RJ (2007) Endoplasmic stress in pancreatic beta cells reveals the early reticulum stress and oxidative stress: a vicious cycle or degradation of insulin mRNAs. Diabetologia 50, 1006– a double-edged sword? Antioxid Redox Signal 9, 2277– 1014. 2293. 56 Malo A, Kruger€ B, Goke€ B & Kubisch CH (2013) 4- 70 Cao SS & Kaufman RJ (2014) Endoplasmic phenylbutyric acid reduces endoplasmic reticulum reticulum stress and oxidative stress in cell fate stress, trypsin activation, and acinar cell apoptosis decision and human disease. Antioxid Redox Signal while increasing secretion in rat pancreatic acini. 21, 396–413. Pancreas 42,92–101. 71 Montibeller L & de Belleroche J (2018) Amyotrophic 57 Lee YY, Hong SH, Lee YJ, Chung SS, Jung HS, Park lateral sclerosis (ALS) and Alzheimer’s disease (AD) SG & Park KS (2010) Tauroursodeoxycholate are characterised by differential activation of ER stress (TUDCA), chemical chaperone, enhances function of pathways: focus on UPR target genes. Cell Stress islets by reducing ER stress. Biochem Biophys Res Chaperones. https://doi.org/10.1007/s12192-018-0897-y Commun 397, 735–739. 72 Bertolotti A, Zhang Y, Hendershot LM, Harding HP 58 Ozcan U, Yilmaz E, Ozcan L, Furuhashi M, & Ron D (2000) Dynamic interaction of BiP and ER Vaillancourt E, Smith RO, Gorg€ un€ CZ & Hotamisligil stress transducers in the unfolded-protein response. GS (2006) Chemical chaperones reduce ER stress and Nat Cell Biol 2, 326–332. restore glucose homeostasis in a mouse model of type 73 Shen J, Chen X, Hendershot L & Prywes R (2002) ER 2 diabetes. Science 313, 1137–1140. stress regulation of ATF6 localization by dissociation 59 Lepock JR (2005) How do cells respond to their of BiP/GRP78 binding and unmasking of Golgi thermal environment? Int J Hyperthermia 21, 681–687. localization signals. Dev Cell 3,99–111. 60 Bettaieb A & Averill-Bates DA (2005) 74 Kopp MC, Nowak PR, Larburu N, Adams CJ & Ali Thermotolerance induced at a mild temperature of MMU (2018) In vitro FRET analysis of IRE1 and BiP 40°C protects cells against heat shock-induced association and dissociation upon endoplasmic apoptosis. J Cell Physiol 205,47–57. reticulum stress. Elife 7, e30257. 61 Liu Y, Sakamoto H, Adachi M, Zhao S, Ukai W, 75 Amin-Wetzel N, Saunders RA, Kamphuis MJ, Rato Hashimoto E, Hareyama M, Ishida T, Imai K & C, Preissler S, Harding HP & Ron D (2017) A J- Shinomura Y (2012) Heat stress activates ER stress Protein co-chaperone recruits BiP to monomerize signals which suppress the heat shock response, an IRE1 and repress the unfolded protein response. Cell effect occurring preferentially in the cortex in rats. Mol 171, 1625–1637.e13. Biol Rep 39, 3987–3993. 76 Sepulveda D, Rojas-Rivera D, Rodrıguez DA, 62 Liu X, Wang M, Chen H, Guo Y, Ma F, Shi F, Bi Y Groenendyk J, Kohler€ A, Lebeaupin C, Ito S, Urra H, & Li Y (2013) Hypothermia protects the brain from Carreras-Sureda A, Hazari Y et al. (2018) Interactome transient global ischemia/reperfusion by attenuating screening identifies the ER luminal chaperone Hsp47 endoplasmic reticulum response-induced apoptosis as a regulator of the unfolded protein response through CHOP. PLoS ONE 8, e53431. transducer IRE1a. Mol Cell 69, 238–252.e7. 63 Mollereau B (2015) Cooling-induced ER stress is good 77 Ye J, Rawson RB, Komuro R, Chen X, Dave UP, for your brain. EBioMedicine 2, 482–483. Prywes R, Brown MS & Goldstein JL (2000) ER stress 64 Zeeshan H, Lee G, Kim H-R & Chae H-J (2016) induces cleavage of membrane-bound ATF6 by the Endoplasmic reticulum stress and associated ROS. Int same proteases that process SREBPs. Mol Cell 6, J Mol Sci 17, 327. 1355–1364.

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 265 Federation of European Biochemical Societies.

38 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

78 Carrara M, Prischi F & Ali MMU (2013) UPR signal initiates mRNA splicing in the unfolded protein activation by luminal sensor domains. Int J Mol Sci response. Cell 90, 1031–1039. 14, 6454–6466. 91 Yoshida H, Matsui T, Yamamoto A, Okada T & 79 Tirasophon W, Welihinda AA & Kaufman RJ (1998) Mori K (2001) XBP1 mRNA is induced by ATF6 A stress response pathway from the endoplasmic and spliced by IRE1 in response to ER stress to reticulum to the nucleus requires a novel bifunctional produce a highly active transcription factor. Cell 107, protein kinase/endoribonuclease (Ire1p) in mammalian 881–891. cells. Genes Dev 12, 1812–1824. 92 Greer CL, Peebles CL, Gegenheimer P & Abelson J 80 Wang XZ, Harding HP, Zhang Y, Jolicoeur EM, (1983) Mechanism of action of a yeast RNA ligase in Kuroda M & Ron D (1998) Cloning of mammalian tRNA splicing. Cell 32, 537–546. Ire1 reveals diversity in the ER stress responses. 93 Sidrauski C, Cox JS & Walter P (1996) tRNA ligase is EMBO J 17, 5708–5717. required for regulated mRNA splicing in the unfolded 81 Iwawaki T, Hosoda A, Okuda T, Kamigori Y, protein response. Cell 87, 405–413. Nomura-Furuwatari C, Kimata Y, Tsuru A & Kohno 94 Schwer B, Sawaya R, Ho CK & Shuman S (2004) K (2001) Translational control by the ER Portability and fidelity of RNA-repair systems. Proc transmembrane kinase/ribonuclease IRE1 under ER Natl Acad Sci USA 101, 2788–2793. stress. Nat Cell Biol 3, 158–165. 95 Tanaka N, Meineke B & Shuman S (2011) RtcB, a 82 Bertolotti A, Wang X, Novoa I, Jungreis R, novel RNA ligase, can catalyze tRNA splicing and Schlessinger K, Cho JH, West AB & Ron D (2001) HAC1 mRNA splicing in vivo. J Biol Chem 286, Increased sensitivity to dextran sodium sulfate 30253–30257. colitis in IRE1b-deficient mice. J Clin Invest 107, 96 Baltz AG, Munschauer M, Schwanhausser€ B, Vasile 585–593. A, Murakawa Y, Schueler M, Youngs N, Penfold- 83 Martino MB, Jones L, Brighton B, Ehre C, Abdulah Brown D, Drew K, Milek M et al. (2012) The L, Davis CW, Ron D, O’Neal WK & Ribeiro CMP mRNA-bound proteome and its global occupancy (2013) The ER stress transducer IRE1b is required for profile on protein-coding transcripts. Mol Cell 46, airway epithelial mucin production. Mucosal Immunol 674–690. 6, 639–654. 97 Lu Y, Liang FX & Wang X (2014) A synthetic 84 Iwawaki T, Akai R, Yamanaka S & Kohno K (2009) biology approach identifies the mammalian UPR RNA Function of IRE1 alpha in the placenta is essential for ligase RtcB. Mol Cell 55, 758–770. placental development and embryonic viability. Proc 98 Liou HC, Boothby MR, Finn PW, Davidon R, Natl Acad Sci USA 106, 16657–16662. Nabavi N, Zeleznik-Le NJ, Ting JP & Glimcher LH 85 Bouchecareilh M, Higa A, Fribourg S, Moenner M & (1990) A new member of the leucine zipper class of Chevet E (2011) Structure of the Ire1 proteins that binds to the HLA DR alpha promoter. autophosphorylation complex and implications for the Science 247, 1581–1584. unfolded protein response. FASEB J 25, 3115– 99 Yoshimura T, Fujisawa J & Yoshida M (1990) 3129. Multiple cDNA clones encoding nuclear proteins that 86 Sanches M, Duffy NM, Talukdar M, Thevakumaran bind to the tax-dependent enhancer of HTLV-1: all N, Chiovitti D, Canny MD, Lee K, Kurinov I, contain a leucine zipper structure and basic amino acid Uehling D, Al-awar R et al. (2014) Structure and domain. EMBO J 9, 2537–2542. mechanism of action of the hydroxy-aryl-aldehyde 100 Nojima H, Leem Sh, Araki H, Sakai A, Nakashima class of IRE1 endoribonuclease inhibitors. Nat N, Kanaoka Y & Ono Y (1994) Hac1: a novel yeast Commun 5, 4202. bZIP protein binding to the CRE motif is a multicopy 87 Prischi F, Nowak PR, Carrara M & Ali MMU (2014) suppressor for cdcW mutant of Schizosaccharomyces Phosphoregulation of Ire1 RNase splicing activity. Nat pombe. Nucleic Acids Res 22, 5279–5288. Commun 5, 3554. 101 Cox JS & Walter P (1996) A novel mechanism for 88 Urano F, Wang X-Z, Bertolotti A, Zhang Y, Chung regulating activity of a transcription factor that P, Harding HP & Ron D (2000) Coupling of stress in controls the unfolded protein response. Cell 87, 391– the endoplasmic reticulum to activation of JNK 404. protein kinases by transmembrane protein kinase 102 Calfon M, Zeng H, Urano F, Till JH, Hubbard SR, IRE1. Science 287, 664–666. Harding HP, Clark SG & Ron D (2002) IRE1 couples 89 Bork P & Sander C (1993) A hybrid protein kinase- endoplasmic reticulum load to secretory capacity by RNase in an interferon-induced pathway? FEBS Lett processing the XBP-1 mRNA. Nature 415,92–96. 334, 149–152. 103 Chen CY, Malchus NS, Hehn B, Stelzer W, Avci D, 90 Sidrauski C & Walter P (1997) The transmembrane Langosch D & Lemberg MK (2014) Signal peptide kinase Ire1p is a site-specific endonuclease that peptidase functions in ERAD to cleave the unfolded

266 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

39 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

protein response regulator XBP1u. EMBO J 33, 2492– MF, Park SW et al. (2016) Inﬂammation improves 2506. glucose homeostasis through IKKb-XBP1s interaction. 104 Yanagitani K, Imagawa Y, Iwawaki T, Hosoda A, Cell 167, 1052–1066.e18. Saito M, Kimata Y & Kohno K (2009) 117 Lee J, Sun C, Zhou Y, Lee J, Gokalp D, Herrema H, Cotranslational Targeting of XBP1 protein to the Park SW, Davis RJ & Ozcan U (2011) p38 MAPK– membrane promotes cytoplasmic splicing of its own mediated regulation of Xbp1s is crucial for glucose mRNA. Mol Cell 34, 191–200. homeostasis. Nat Med 17, 1251–1260. 105 Yanagitani K, Kimata Y, Kadokura H & Kohno K 118 Yingfeng D, Zhao WV, Caroline T, Ningguo G, (2011) Translational pausing ensures membrane William HL, Anwarul F, Joyce RJ, Guosheng L, Jin targeting and cytoplasmic splicing of XBP1u mRNA. Y, Mark LA et al. (2013) The Xbp1s/GalE axis links Science 331, 586–589. ER stress to postprandial hepatic metabolism. J Clin 106 Kanda S, Yanagitani K, Yokota Y, Esaki Y & Kohno Invest 123, 455–468. K (2016) Autonomous translational pausing is 119 Ozcan€ U, Cao Q, Yilmaz E, Lee A-H, Iwakoshi NN, required for XBP1u mRNA recruitment to the ER via Ozdelen€ E, Tuncman G, Gorg€ un€ C, Glimcher Laurie the SRP pathway. Proc Natl Acad Sci 113, E5886– H & Hotamisligil GS (2004) Endoplasmic reticulum E5895. stress links obesity, insulin action, and type 2 diabetes. 107 Chalmers F, Sweeney B, Cain K & Bulleid NJ (2017) Science 306, 457–461. Inhibition of IRE1a-mediated XBP1 mRNA cleavage 120 Akiyama M, Liew CW, Lu S, Hu J, Martinez R, by XBP1 reveals a novel regulatory process during the Hambro B, Kennedy RT & Kulkarni RN (2013) X- unfolded protein response. Wellcome Open Res 2, 36. box binding protein 1 is essential for insulin regulation 108 He Y, Sun S, Sha H, Liu Z, Yang L, Xue Z, Chen H of pancreatic a-cell function. Diabetes 62, 2439–2449. & Qi L (2010) Emerging roles for XBP1, a sUPeR 121 Liu Y, Adachi M, Zhao S, Hareyama M, Koong AC, transcription factor. Gene Expr 15,13–25. Luo D & Rando TA (2010) Preventing oxidative stress 109 Travers KJ, Patil CK, Wodicka L, Lockhart DJ, a new role for XBP1. Cell 16, 847–857. Weissman JS & Walter P (2000) Functional and 122 Tao R, Chen H, Gao C, Xue P, Yang F, Han J-DJ, genomic analyses reveal an essential coordination Zhou B & Chen Y-G (2011) Xbp1-mediated histone between the unfolded protein response and ER- H4 deacetylation contributes to DNA double-strand associated degradation. Cell 101, 249–258. break repair in yeast. Cell Res 21, 1619–1633. 110 Iwakoshi NN, Lee A-H & Glimcher LH (2003) The 123 Wu J & Kaufman RJ (2006) From acute ER stress to X-box binding protein-1 transcription factor is physiological roles of the Unfolded Protein Response. required for plasma cell differentiation and the Cell Death Differ 13, 374–384. unfolded protein response. Immunol Rev 194,29–38. 124 Blais A, Tsikitis M, Acosta-Alvear D, Sharan R, 111 Sriburi R, Jackowski S, Mori K & Brewer JW (2004) Kluger Y & Dynlacht BD (2005) An initial blueprint XBP1: a link between the unfolded protein response, for myogenic differentiation. Genes Dev 19, 553–569. lipid biosynthesis, and biogenesis of the endoplasmic 125 Reimold AM, Iwakoshi NN, Manis J, Vallabhajosyula reticulum. J Cell Biol 167,35–41. P, Szomolanyi-Tsuda E, Gravallese EM, Friend D, 112 Korchak HM (2008) Regulation of hepatic lipogenesis. Grusby MJ, Alt F & Glimcher LH (2001) Plasma cell Tufts Folia Med 8, 134–143. differentiation requires the transcription factor XBP-1. 113 So JS, Hur KY, Tarrio M, Ruda V, Frank- Nature 412, 300–307. Kamenetsky M, Fitzgerald K, Koteliansky V, 126 Lee A-H, Chu GC, Iwakoshi NN & Glimcher LH Lichtman AH, Iwawaki T, Glimcher LH et al. (2012) (2005) XBP-1 is required for biogenesis of cellular Silencing of lipid metabolism genes through IRE1a- secretory machinery of exocrine glands. EMBO J 24, mediated mRNA decay lowers plasma lipids in mice. 4368–4380. Cell Metab 16, 487–499. 127 Huh WJ, Esen E, Geahlen JH, Bredemeyer AJ, Lee A, 114 Zhou Y, Lee J, Reno CM, Sun C, Park SW, Chung J, Shi G, Konieczny SF, Glimcher LH & Mills JC (2010) Lee J, Fisher SJ, White MF, Biddinger SB et al. XBP1 controls maturation of gastric zymogenic cells (2011) Regulation of glucose homeostasis through a by induction of MIST1 and expansion of the rough XBP-1–FoxO1 interaction. Nat Med 17, 356–365. endoplasmic reticulum. Gastroenterology 139, 2038– 115 Park SW, Herrema H, Salazar M, Cakir I, Cabi S, 2049. Basibuyuk Sahin F, Chiu YH, Cantley LC & Ozcan U 128 Sha H, He Y, Chen H, Wang C, Zenno A, Shi H, (2014) BRD7 regulates XBP1s’ activity and glucose Yang X & Zhang X (2009) The IRE1a-XBP1 pathway homeostasis through its interaction with the regulatory of the unfolded protein response is required for subunits of PI3K. Cell Metab 20,73–84. adipogenesis. Cell Metab 9, 556–564. 116 Liu J, Ibi D, Taniguchi K, Lee J, Herrema H, 129 Masaki T, Yoshida M & Noguchi S (1999) Targeted Akosman B, Mucka P, Salazar Hernandez MA, Uyar disruption of CRE-Binding factor TREB5 gene leads

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 267 Federation of European Biochemical Societies.

40 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

to cellular necrosis in cardiac myocytes at the 141 Hollien J & Weissman JS (2006) Decay of embryonic stage. Biochem Biophys Res Commun 261, endoplasmic reticulum-localized mRNAs 350–356. during the unfolded protein response. Science 313, 130 Reimold AM, Etkin A, Clauss I, Perkins A, Friend 104–107. DS, Zhang J, Horton HF, Scott A, Orkin SH, Byrne 142 Han D, Lerner AG, Vande WL, Upton J-P, Xu W, MC et al. (2000) An essential role in liver development Hagen A, Backes BJ, Oakes SA & Papa FR (2009) for transcription factor XBP-1. Genes Dev 14, 152– IRE1a kinase activation modes control alternate 157. endoribonuclease outputs to determine divergent cell 131 Sone M, Zeng X, Larese J & Ryoo HD (2013) A fates. Cell 138, 562–575. modified UPR stress sensing system reveals a novel 143 Kimmig P, Diaz M, Zheng J, Williams CC, Lang A, tissue distribution of IRE1/XBP1 activity during Aragon T, Li H & Walter P (2012) The unfolded normal Drosophila development. Cell Stress protein response in fission yeast modulates stability of Chaperones 18, 307–319. select mRNAs to maintain protein homeostasis. Elife 132 Ono SJ, Liout H, Davidont R, Strominger JL & 2012,1–20. Glimchert LH (1991) Human X-box-binding protein 1 144 Mishiba K-0049, Nagashima Y, Suzuki E, Hayashi N, is required for the transcription of a subset of human Ogata Y, Shimada Y & Koizumi N (2013) Defects in class II major histocompatibility genes and forms a IRE1 enhance cell death and fail to degrade mRNAs heterodimer with c-fos. Proc Natl Acad Sci USA 88, encoding secretory pathway proteins in the 4309–4312. Arabidopsis unfolded protein response. Proc Natl 133 Ding L, Yan J, Zhu J, Zhong H, Lu Q, Wang Z, Acad Sci 110, 5713–5718. Huang C & Ye Q (2003) Ligand-independent 145 Hayashi S, Wakasa Y, Ozawa K & Takaiwa F (2016) activation of estrogen receptor a by XBP-1. Nucleic Characterization of IRE1 ribonuclease-mediated Acids Res 31, 5266–5274. mRNA decay in plants using transient expression 134 Ravasi T, Suzuki H, Cannistraci CV, Katayama S, analyses in rice protoplasts. New Phytol 210, 1259– Bajic VB, Tan K, Akalin A, Schmeier S, Kanamori- 1268. Katayama M, Bertin N et al. (2010) An atlas of 146 Oikawa D, Tokuda M, Hosoda A & Iwawaki T combinatorial transcriptional regulation in mouse and (2010) Identification of a consensus element recognized man. Cell 140, 744–752. and cleaved by IRE1a. Nucleic Acids Res 38, 6265– 135 Reinke AW, Baek J, Ashenberg O & Keating AE 6273. (2013) Networks of bZIP protein-protein interactions 147 Iqbal J, Dai K, Seimon T, Jungreis R, Oyadomari M, diversified over a billion years of evolution. Science Kuriakose G, Ron D, Tabas I & Hussain MM (2008) 340, 730–734. IRE1b Inhibits chylomicron production by selectively 136 Chen X, Iliopoulos D, Zhang Q, Tang Q, Greenblatt degrading MTP mRNA. Cell Metab 7, 445–455. MB, Hatziapostolou M, Lim E, Tam WL, Ni M, 148 Imagawa Y, Hosoda A, Sasaka S, Tsuru A & Kohno Chen Y et al. (2014) XBP1 promotes triple-negative K (2008) RNase domains determine the functional breast cancer by controlling the HIF1a pathway. difference between IRE1a and IRE1b. FEBS Lett 582, Nature 508, 103–107. 656–660. 137 Hollien J, Lin JH, Li H, Stevens N, Walter P & 149 Shi Y, Vattem KM, Sood R, An J, Liang J, Stramm L Weissman JS (2009) Regulated Ire1-dependent decay & Wek RC (1998) Identification and characterization of messenger RNAs in mammalian cells. J Cell Biol of pancreatic eukaryotic initiation factor 2 alpha- 186, 323–331. subunit kinase, PEK, involved in translational control. 138 Maurel M, Chevet E, Tavernier J & Gerlo S (2014) Mol Cell Biol 18, 7499–7509. Getting RIDD of RNA: IRE1 in cell fate regulation. 150 Harding HP, Zhang Y & Ron D (1999) Protein Trends Biochem Sci 39, 245–254. translation and folding are coupled by an 139 Lhomond S, Avril T, Dejeans N, Voutetakis K, endoplasmic-reticulum-resident kinase. Nature 397, Doultsinos D, McMahon M, Pineau R, Obacz J, 271–274. Papadodima O, Jouan F et al. (2018) Dual IRE1 151 McQuiston A & Diehl JA (2017) Recent insights into RNase functions dictate glioblastoma development. PERK-dependent signaling from the stressed EMBO Mol Med 10, e7929. https://doi.org/10.15252/ endoplasmic reticulum. F1000Research 6, 1897. emmm.201707929 152 Lloyd MA, Osborne JC, Safer B, Powell GM & 140 Tirasophon W, Lee K, Callaghan B, Welihinda A & Merrick WC (1980) Characteristics of eukaryotic Kaufman RJ (2000) The endoribonuclease activity of initiation factor 2 and its subunits. J Biol Chem 255, mammalian IRE1 autoregulates its mRNA and is 1189–1193. required for the unfolded protein response. Genes Dev 153 Ernst H, Duncan RF & Hershey JWB (1987) Cloning 14, 2725–2736. and sequencing of complementary DNAs encoding the

268 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

41 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

alpha-subunit of translational initiation factor-Eif-2 - 165 Barbosa C, Peixeiro I & Romao~ L (2013) Gene characterization of the protein and its messenger expression regulation by upstream open reading RNA. J Biol Chem 262, 1206–1212. frames and human disease. PLoS Genet 9,1–12. 154 Adams SL, Safer B & Anderson WFMW (1975) 166 Fornace AJ, Alamo I & Hollander MC (1988) DNA Eukaryotic initiation complex formation. Evidence for damage-inducible transcripts in mammalian cells. Proc two distinct pathways. J Biol Chem 250, 9083–9089. Natl Acad Sci USA 85, 8800–8804. 155 Rowlands AG, Panniers R & Henshaw EC (1988) The 167 Fornace AJ, Nebert DW, Hollander MC, Luethy JD, catalytic mechanism of guanine nucleotide exchange Papathanasiou M, Fargnoli J & Holbrook NJ (1989) factor action and competitive inhibition by Mammalian genes coordinately regulated by growth phosphorylated eukaryotic initiation factor 2. J Biol arrest signals and DNA-damaging agents. Mol Cell Chem 263, 5526–5533. Biol 9, 4196–4203. 156 Harding HP, Zhang Y, Bertolotti A, Zeng H & Ron 168 Lee YY, Cevallos RC & Jan E (2009) An upstream D (2000) Perk is essential for translational regulation open reading frame regulates translation of GADD34 and cell survival during the unfolded protein response. during cellular stresses that induce eIF2 Mol Cell 5, 897–904. phosphorylation. J Biol Chem 284, 6661–6673. 157 Hai T, Liu F, Coukos WJ & Green MR (1989) 169 Ma Y & Hendershot LM (2003) Delineation of a Transcription factor ATF cDNA clones: an extensive negative feedback regulatory loop that controls family of leucine zipper proteins able to selectively protein translation during endoplasmic reticulum form DNA-binding heterodimers. Genes Dev 3, 2083– stress. J Biol Chem 278, 34864–34873. 2090. 170 Marciniak SJ, Yun CY, Oyadomari S, Novoa I, 158 Harding HP, Novoa I, Zhang Y, Zeng H, Wek R, Zhang Y, Jungreis R, Nagata K, Harding HP & Ron Schapira M & Ron D (2000) Regulated translation D (2004) CHOP induces death by promoting protein initiation controls stress-induced gene expression in synthesis and oxidation in the stressed endoplasmic mammalian cells. Mol Cell 6, 1099–1108. reticulum. Genes Dev 18, 3066–3077. 159 Vallejo M, Ron D, Miller CP & Habener JF (1993) C/ 171 Connor JH, Weiser DC, Li S, John M, Li SHI & ATF, a member of the activating transcription factor Hallenbeck JM (2001) Growth arrest and DNA family of DNA-binding proteins, dimerizes with damage-inducible protein GADD34 assembles a novel CAAT/enhancer-binding proteins and directs their signaling complex containing Protein Phosphatase 1 binding to cAMP response elements. Proc Natl Acad and inhibitor 1. Mol Cell Biol 21, 6841–6850. Sci USA 90, 4679–4683. 172 Novoa I, Zeng H, Harding HP & Ron D (2001) 160 Fawcett TW, Martindale JL, Guyton KZ, Hai T & Feedback inhibition of the unfolded protein response Holbrook NJ (1999) Complexes containing activating by GADD34-mediated dephosphorylation of transcription factor (ATF)/cAMP-responsive-element- eIF2alpha. J Cell Biol 153, 1011–1022. binding protein (CREB) interact with the CCAAT/ 173 Harding HP, Zhang Y, Zeng H, Novoa I, Lu PD, enhancer-binding protein (C/EBP)–ATF composite site Calfon M, Sadri N, Yun C, Popko B, Paules R et al. to regulate Gadd153 expression during the stress (2003) An integrated stress response regulates amino response. Biochem J 339, 135–141. acid metabolism and resistance to oxidative stress. 161 Han J, Back SH, Hur J, Lin Y, Gildersleeve R, Shan Mol Cell 11, 619–633. J, Yuan CL, Krokowski D, Wang S, Hatzoglou M 174 Quiros PM, Prado MA, Zamboni N, D’Amico D, et al. (2013) ER-stress-induced transcriptional Williams RW, Finley D, Gygi SP & Auwerx J (2017) regulation increases protein synthesis leading to cell Multi-omics analysis identiﬁes ATF4 as a key death. Nat Cell Biol 15, 481–490. regulator of the mitochondrial stress response in 162 Hiramatsu N, Messah C, Han J, LaVail MM, Kaufman mammals. J Cell Biol 216, 2027–2045. RJ & Lin JH (2014) Translational and posttranslational 175 Rajesh K, Krishnamoorthy J, Kazimierczak U, regulation of XIAP by eIF2a and ATF4 promotes ER Tenkerian C, Papadakis AI, Wang S, Huang S & stress-induced cell death during the unfolded protein Koromilas AE (2015) Phosphorylation of the response. Mol Biol Cell 25, 1411–1420. translation initiation factor eIF2a at serine 51 163 Chen YJ, Tan BCM, Cheng YY, Chen JS & Lee SC determines the cell fate decisions of Akt in response to (2009) Differential regulation of CHOP translation by oxidative stress. Cell Death Dis 6, e1591. phosphorylated eIF4E under stress conditions. Nucleic 176 Lee IC, Ho XY, George SE, Goh CW, Sundaram Acids Res 38, 764–777. JR, Pang KKL, Luo W, Yusoff P, Sze NSK & 164 Palam LR, Baird TD & Wek RC (2011) Shenolikar S (2017) Oxidative stress promotes SIRT1 Phosphorylation of eIF2 facilitates ribosomal bypass recruitment to the GADD34/PP1a complex to of an inhibitory upstream ORF to enhance CHOP activate its deacetylase function. Cell Death Differ 25, translation. J Biol Chem 286, 10939–10949. 255–267.

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 269 Federation of European Biochemical Societies.

42 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

177 Chen D, Fan Z, Rauh M, Buchfelder M, Eyupoglu IY 190 Gorman AM, Healy SJ, Jager R & Samali A (2012) & Savaskan N (2017) ATF4 promotes angiogenesis Stress management at the ER: regulators of ER stress- and neuronal cell death and confers ferroptosis in a induced apoptosis. Pharmacol Ther 134, 306–316. xCT-dependent manner. Oncogene 36, 5593–5608. 191 Ron D & Walter P (2007) Signal integration in the 178 Wu Z, Li M, Zheng W, Hu Q, Cheng Z & Guo F endoplasmic reticulum unfolded protein response. Nat (2017) Silencing of both ATF4 and PERK inhibits cell Rev Mol Cell Biol 8, 519–529. cycle progression and promotes the apoptosis of 192 Hoyer-Hansen M & Jaattela M (2007) Connecting differentiating chondrocytes. Int J Mol Med 40, 101– endoplasmic reticulum stress to autophagy by 111. unfolded protein response and calcium. Cell Death 179 Ishizawa J, Kojima K, Chachad D, Ruvolo P, Ruvolo Differ 14, 1576–1582. V, Jacamo RO, Borthakur G, Mu H, Zeng Z, Tabe Y 193 Cheng X, Liu H, Jiang CC, Fang L, Chen C, Zhang et al. (2016) ATF4 induction through an atypical XD & Jiang ZW (2014) Connecting endoplasmic integrated stress response to ONC201 triggers p53- reticulum stress to autophagy through IRE1/JNK/ independent apoptosis in hematological malignancies. beclin-1 in breast cancer cells. Int J Mol Med 34, 772– Sci Signal 9, ra17. 781. 180 Iurlaro R, Puschel€ F, Leon-Annicchiarico CL, 194 Giorgi C, Missiroli S, Patergnani S, Duszynski J, O’Connor H, Martin SJ, Palou-Gramon D, Lucendo Wieckowski MR & Pinton P (2015) Mitochondria- E&Munoz-Pinedo~ C (2017) Glucose deprivation associated membranes: composition, molecular induces ATF4-mediated apoptosis through TRAIL mechanisms, and physiopathological implications. death receptors. Mol Cell Biol 37, e00479–16. https:// Antioxid Redox Signal 22, 995–1019. doi.org/10.1128/MCB.00479-16 195 Santel A & Fuller MT (2001) Control of 181 Zhu C, Johansen FE & Prywes R (1997) Interaction mitochondrial morphology by a human mitofusin. J of ATF6 and serum response factor. Mol Cell Biol 17, Cell Sci 114, 867–874. 4957–4966. 196 Munoz~ JP, Ivanova S, Sanchez-Wandelmer J, 182 Yamamoto K, Sato T, Matsui T, Sato M, Okada T, Martınez-Cristobal P, Noguera E, Sancho A, Dıaz- Yoshida H, Harada A & Mori K (2007) Transcriptional Ramos A, Hernandez-Alvarez MI, Sebastian D, induction of mammalian ER quality control proteins is Mauvezin C et al. (2014) Erratum: Mfn2 modulates mediated by single or combined action of ATF6alpha the UPR and mitochondrial function via repression of and XBP1. Dev Cell 13, 365–376. PERK (EMBO Journal 32 (2348-2361) https://doi.org/ 183 Asada R, Kanemoto S, Kondo S, Saito A & Imaizumi 10.1038/emboj.2013.168). EMBO J 33, 171. K (2011) The signalling from endoplasmic reticulum- 197 Verfaillie T, Rubio N, Garg AD, Bultynck G, Rizzuto resident bZIP transcription factors involved in diverse R, Decuypere JP, Piette J, Linehan C, Gupta S, Samali cellular physiology. J Biochem 149, 507–518. A et al. (2012) PERK is required at the ER- 184 McMahon M, Samali A & Chevet E (2017) mitochondrial contact sites to convey apoptosis after Regulation of the unfolded protein response by ROS-based ER stress. Cell Death Differ 19, 1880–1891. noncoding RNA. Am J Physiol Cell Physiol 313, 198 Mori T, Hayashi T, Hayashi E & Su TP (2013) C243–C254. Sigma-1 receptor chaperone at the er-mitochondrion 185 Oslowski CM & Urano F (2011) Measuring ER stress interface mediates the mitochondrion-er-nucleus and the unfolded protein response using mammalian signaling for cellular survival. PLoS ONE 8, e76941. tissue culture system. Methods Enzymol 490,71–92. 199 Lisbona F, Rojas-Rivera D, Thielen P, Zamorano S, 186 Yoshida H, Matsui T, Hosokawa N, Kaufman RJ, Todd D, Martinon F, Glavic A, Kress C, Lin JH, Nagata K & Mori K (2003) A time-dependent phase Walter P et al. (2009) BAX inhibitor-1 is a negative shift in the mammalian unfolded protein response. Dev regulator of the ER stress sensor IRE1alpha. Mol Cell Cell 4, 265–271. 33, 679–691. 187 Sano R & Reed JC (2013) ER stress-induced cell death 200 Vannuvel K, Renard P, Raes M & Arnould T (2013) mechanisms. Biochim Biophys Acta 1833, 3460–3470. Functional and morphological impact of ER stress on 188 Bartoszewska S, Kochan K, Madanecki P, Piotrowski mitochondria. J Cell Physiol 228, 1802–1818. A, Ochocka R, Collawn JF & Bartoszewski R (2013) 201 Bravo R, Vicencio JM, Parra V, Troncoso R, Munoz Regulation of the unfolded protein response by JP, Bui M, Quiroga C, Rodriguez AE, Verdejo HE, microRNAs. Cell Mol Biol Lett 18, 555–578. Ferreira J et al. (2011) Increased ER-mitochondrial 189 Wortel IMN, van der Meer LT, Kilberg MS & van coupling promotes mitochondrial respiration and Leeuwen FN (2017) Surviving stress: modulation of bioenergetics during early phases of ER stress. J Cell ATF4-mediated stress responses in normal Sci 124, 2511. and malignant cells. Trends Endocrinol Metab 28, 202 Cullinan SB & Diehl JA (2004) PERK-dependent 794–806. activation of Nrf2 contributes to redox homeostasis

270 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

43 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

and cell survival following endoplasmic reticulum induced hypertriglyceridemia in mice. J Biol Chem stress. J Biol Chem 279, 20108–20117. 288, 25350–25361. 203 Wang C, Li H, Meng Q, Du Y, Xiao F, Zhang Q, Yu 215 Wang S, Chen Z, Lam V, Han J, Hassler J, Finck J, Li K, Chen S, Huang Z et al. (2014) ATF4 BN, Davidson NO & Kaufman RJ (2012) IRE1a- deficiency protects hepatocytes from oxidative stress XBP1s induces PDI expression to increase MTP via inhibiting CYP2E1 expression. J Cell Mol Med 18, activity for hepatic VLDL assembly and lipid 80–90. homeostasis. Cell Metab 16, 473–486. 204 Li G, Mongillo M, Chin KT, Harding H, Ron D, 216 Lee AH, Scapa EF, Cohen DE & Glimcher LH (2008) Marks AR & Tabas I (2009) Role of ERO1-a- Regulation of hepatic lipogenesis by the transcription mediated stimulation of inositol 1,4,5-triphosphate factor XBP1. Science (80-) 320, 1492–1496. receptor activity in endoplasmic reticulum stress- 217 Yi H, Gu C, Li M, Zhang Z, Li Q, Feng J, Zhou J & induced apoptosis. J Cell Biol 186, 783–792. Du J (2017) PERK/eIF2a contributes to changes of 205 Eletto D, Chevet E, Argon Y & Appenzeller-Herzog C insulin signaling in HepG2 cell induced by intermittent (2014) Redox controls UPR to control redox. J Cell hypoxia. Life Sci 181,17–22. Sci 127, 3649–3658. 218 Hosogai N, Fukuhara A, Oshima K, Miyata Y, 206 Appenzeller-Herzog C & Hall MN (2012) Bidirectional Tanaka S, Segawa K, Furukawa S, Tochino Y, crosstalk between endoplasmic reticulum stress and Komuro R, Matsuda M et al. (2007) Adipose tissue mTOR signaling. Trends Cell Biol 22, 274–282. hypoxia in obesity and its impact on adipocytokine 207 Patricia F, Julien B, Justine L, Patrick M, Jo€elle F, dysregulation. Diabetes 56, 901–911. Cecile V, Thomas W, Serge M, Colette R, Jean-Yves S 219 Kim B, Kim MS & Hyun CK (2017) Syringin et al. (2017) mTOR inhibitors activate PERK attenuates insulin resistance via adiponectin-mediated signaling and favor viability of gastrointestinal suppression of low-grade chronic inflammation and neuroendocrine cell lines. Oncotarget 8, 20974–20987. ER stress in high-fat diet-fed mice. Biochem Biophys 208 Tenkerian C, Krishnamoorthy J, Mounir Z, Res Commun 488,40–45. Kazimierczak U, Khoutorsky A, Staschke KA, Kristof 220 Lipson KL, Fonseca SG, Ishigaki S, Nguyen LX, Foss AS, Wang S, Hatzoglou M & Koromilas AE (2015) E, Bortell R, Rossini AA & Urano F (2006) mTORC2 balances AKT activation and eIF2a serine Regulation of insulin biosynthesis in pancreatic beta 51 phosphorylation to promote survival under stress. cells by an endoplasmic reticulum-resident protein Mol Cancer Res 13, 1377–1388. kinase IRE1. Cell Metab 4, 245–254. 209 Feng B, Yao PM, Li Y, Devlin CM, Zhang D, 221 Lee A-H, Heidtman K, Hotamisligil GS & Glimcher Harding HP, Sweeney M, Rong JX, Kuriakose G, LH (2011) Dual and opposing roles of the unfolded Fisher EA et al. (2003) The endoplasmic reticulum is protein response regulated by IRE1alpha and XBP1 in the site of cholesterol-induced cytotoxicity in proinsulin processing and insulin secretion. Proc Natl macrophages. Nat Cell Biol 5, 781–792. Acad Sci 108, 8885–8890. 210 Fu S, Yang L, Li P, Hofmann O, Dicker L, Hide W, 222 Seo H-Y, Kim YD, Lee K-M, Min A-K, Kim M-K, Lin X, Watkins SM, Ivanov AR & Hotamisligil GS Kim H-S, Won K-C, Park J-Y, Lee K-U, Choi H-S (2011) Aberrant lipid metabolism disrupts calcium et al. (2008) Endoplasmic reticulum stress-induced homeostasis causing liver endoplasmic reticulum stress activation of activating transcription factor 6 decreases in obesity. Nature 473, 528–531. insulin gene expression via up-regulation of orphan 211 Kaplowitz N, Than TA, Ph D, Shinohara M & Ji C nuclear receptor small heterodimer partner. (2007) Endoplasmic reticulum stress and liver injury. Endocrinology 149, 3832–3841. Semin Liver Dis 27, 367–377. 223 Shao M, Shan B, Liu Y, Deng Y, Yan C, Wu Y, Mao 212 Oyadomari S, Harding HP, Zhang Y, Oyadomari M T, Qiu Y, Zhou Y, Jiang S et al. (2014) Hepatic & Ron D (2008) De-phosphorylation of translation IRE1a regulates fasting-induced metabolic adaptive initiation factor 2a (eIF2a) enhances glucose tolerance programs through the XBP1s–PPARa axis signalling. and attenuates hepato-steatosis in mice. Cell Metab 7, Nat Commun 5, 3528. 520–532. 224 Kim H, Tu HC, Ren D, Takeuchi O, Jeffers JR, 213 Li H, Meng Q, Xiao F, Chen S, Du Y, Yu J, Wang C Zambetti GP, Hsieh JJ & Cheng EH (2009) Stepwise & Guo F (2011) ATF4 deficiency protects mice from activation of BAX and BAK by tBID, BIM, and high-carbohydrate-diet-induced liver steatosis. Biochem PUMA initiates mitochondrial apoptosis. Mol Cell 36, J 438, 283–289. 487–499. 214 Xiao G, Zhang T, Yu S, Lee S, Calabuig-Navarro V, 225 Gomez-Bougie P, Halliez M, Moreau P, Pellat- Yamauchi J, Ringquist S & Dong HH (2013) ATF4 Deceunynck C & Amiot M (2016) Repression of Mcl- protein deficiency protects against high fructose- 1 and disruption of the Mcl-1/Bak interaction in

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 271 Federation of European Biochemical Societies.

44 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

myeloma cells couple ER stress to mitochondrial 237 Galluzzi L, Buque A, Kepp O, Zitvogel L & Kroemer apoptosis. Cancer Lett 383, 204–211. G (2016) Immunogenic cell death in cancer and 226 Li J, Lee B & Lee AS (2006) Endoplasmic reticulum infectious disease. Nat Rev Immunol 17, 97. stress-induced apoptosis: multiple pathways and 238 Garg AD & Agostinis P (2017) Cell death and activation of p53-up-regulated modulator of apoptosis immunity in cancer: from danger signals to mimicry of (PUMA) and NOXA by p53. J Biol Chem 281, 7260– pathogen defense responses. Immunol Rev 280, 126– 7270. 148. 227 Ogata M, Hino S, Saito A, Morikawa K, Kondo S, 239 Fan H, Tang HB, Kang J, Shan L, Song H, Zhu K, Kanemoto S, Murakami T, Taniguchi M, Tanii I, Wang J, Ju G & Wang YZ (2015) Involvement of Yoshinaga K et al. (2006) Autophagy is activated for endoplasmic reticulum stress in the necroptosis of cell survival after endoplasmic reticulum stress. Mol microglia/macrophages after spinal cord injury. Cell Biol 26, 9220–9231. Neuroscience 311, 362–373. 228 Yu L, Alva A, Su H, Dutt P, Freundt E, Welsh S, 240 Linkermann A & Green DR (2014) Necroptosis. N Baehrecke EH & Lenardo MJ (2004) Regulation of an Engl J Med 370, 455–465. ATG7-beclin 1 program of autophagic cell death by 241 Saveljeva S, Mc Laughlin SL, Vandenabeele P, Samali caspase-8. Science (80-.) 304, 1500–1502. A & Bertrand MJ (2015) Endoplasmic reticulum stress 229 Kim I, Shu CW, Xu W, Shiau CW, Grant D, Vasile induces ligand-independent TNFR1-mediated S, Cosford ND & Reed JC (2009) Chemical biology necroptosis in L929 cells. Cell Death Dis 6, e1587. investigation of cell death pathways activated by 242 Pattingre S, Tassa A, Qu X, Garuti R, Liang XH, endoplasmic reticulum stress reveals cytoprotective Mizushima N, Packer M, Schneider MD & Levine B modulators of ASK1. J Biol Chem 284, 1593– (2005) Bcl-2 antiapoptotic proteins inhibit Beclin 1- 1603. dependent autophagy. Cell 122, 927–939. 230 Deng X, Xiao L, Lang W, Gao F, Ruvolo P & May 243 Gandin V, Pellei M, Tisato F, Porchia M, Santini C WS (2001) Novel Role for JNK as a Stress-activated & Marzano C (2012) A novel copper complex Bcl2 Kinase. J Biol Chem 276, 23681–23688. induces paraptosis in colon cancer cells via the 231 Puthalakath H, O’Reilly LA, Gunn P, Lee L, Kelly activation of ER stress signalling. J Cell Mol Med PN, Huntington ND, Hughes PD, Michalak EM, 16, 142–151. McKimm-Breschkin J, Motoyama N et al. (2007) ER 244 Fan J, Long H, Li Y, Liu Y, Zhou W, Li Q, Yin G, stress triggers apoptosis by activating BH3-only Zhang N & Cai W (2013) Edaravone protects against protein Bim. Cell 129, 1337–1349. glutamate-induced PERK/EIF2a/ATF4 integrated 232 Yamaguchi H & Wang H-G (2004) CHOP is involved stress response and activation of caspase-12. Brain Res in endoplasmic reticulum stress-induced apoptosis by 1519,1–8. enhancing DR5 expression in human carcinoma cells. 245 Akamatsu K, Shibata M-A, Ito Y, Sohma Y, Azuma J Biol Chem 279, 45495–45502. H & Otsuki Y (2009) Riluzole induces apoptotic cell 233 Maytin EV, Ubeda M, Lin JC & Habener JF (2001) death in human prostate cancer cells via endoplasmic Stress-Inducible Transcription Factor CHOP/gadd153 reticulum stress. Anticancer Res 29, 2195–2204. Induces Apoptosis in Mammalian Cells via p38 246 Cao J, Dai DL, Yao L, Yu HH, Ning B, Zhang Q, Kinase-Dependent and -Independent Mechanisms. Chen J, Cheng WH, Shen W & Yang ZX (2012) Exp Cell Res 267, 193–204. Saturated fatty acid induction of endoplasmic 234 Wei S-G, Yu Y, Weiss RM & Felder RB (2016) reticulum stress and apoptosis in human liver cells via Endoplasmic reticulum stress increases brain MAPK the PERK/ATF4/CHOP signaling pathway. Mol Cell signaling, inﬂammation and renin-angiotensin system Biochem 364, 115–129. activity and sympathetic nerve activity in heart failure. 247 Du S, Zhou J, Jia Y & Huang K (2010) SelK is a Am J Physiol Heart Circ Physiol 311, H871–H880. novel ER stress-regulated protein and protects HepG2 235 Nolan K, Walter F, Tuffy LP, Poeschel S, Gallagher cells from ER stress agent-induced apoptosis. Arch R, Haunsberger S, Bray I, Stallings RL, Concannon Biochem Biophys 502, 137–143. CG & Prehn JH (2016) Endoplasmic reticulum stress- 248 Li S, Zhao F, Cheng S, Wang X & Hao Y (2013) Uric mediated upregulation of miR-29a enhances sensitivity acid-induced endoplasmic reticulum stress triggers to neuronal apoptosis. Eur J Neurosci 43, 640–652. phenotypic change in rat glomerular mesangial cells. 236 Wu Y, Li X, Jia J, Zhang Y, Li J, Zhu Z, Wang H, Nephrology 18, 682–689. Tang J & Hu J (2018) Transmembrane E3 ligase 249 Apostolou A, Shen Y, Liang Y, Luo J & Fang S RNF183 mediates ER stress-induced apoptosis by (2008) Armet, a UPR-upregulated protein, inhibits cell degrading Bcl-xL. Proc Natl Acad Sci USA 115, proliferation and ER stress-induced cell death. Exp E2762–E2771. Cell Res 314, 2454–2467.

272 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

45 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

250 Morishima N, Nakanishi K, Takenouchi H, Shibata T Murphy D (2014) Transcription factor CREB3L1 & Yasuhiko Y (2002) An endoplasmic reticulum regulates vasopressin gene expression in the rat stress-specific caspase cascade in apoptosis. hypothalamus. J Neurosci 34, 3810–3820. Cytochrome c-independent activation of caspase-9 by 261 Azuma Y, Hagiwara D, Lu W, Morishita Y, Suga H, caspase-12. J Biol Chem 277, 34287–34294. Goto M, Banno R, Sugimura Y, Oyadomari S, Mori 251 Saito A, Ochiai K, Kondo S, Tsumagari K, K et al. (2014) Activating transcription factor 6a is Murakami T, Cavener DR & Imaizumi K (2011) required for the vasopressin neuron system to Endoplasmic reticulum stress response mediated by the maintain water balance under dehydration in male PERK-eIF2(alpha)-ATF4 pathway is involved in mice. Endocrinology 155, 4905–4914. osteoblast differentiation induced by BMP2. J Biol 262 Marwarha G, Claycombe K, Schommer J, Collins D Chem 286, 4809–4818. & Ghribi O (2016) Palmitate-induced endoplasmic 252 Chang YJ, Chen WY, Huang CY, Liu HH & Wei reticulum stress and subsequent C/EBPa homologous PL (2015) Glucose-regulated protein 78 (GRP78) protein activation attenuates leptin and Insulin-like regulates colon cancer metastasis through EMT growth factor 1 expression in the brain. Cell Signal 28, biomarkers and the NRF-2/HO-1 pathway. Tumor 1789–1805. Biol 36, 1859–1869. 263 Marwarha G, Dasari B & Ghribi O (2012) 253 Tanjore H, Cheng D-S, Degryse AL, Zoz DF, Endoplasmic reticulum stress-induced CHOP Abdolrasulnia R, Lawson WE & Blackwell TS (2011) activation mediates the down-regulation of leptin in Alveolar epithelial cells undergo epithelial-to- human neuroblastoma SH-SY5Y cells treated with the mesenchymal transition in response to endoplasmic oysterol 27-hydroxycholesterol. Cell Signal 24, 484– reticulum stress. J Biol Chem 286, 30972–30980. 492. 254 Mo XT, Zhou WC, Cui WH, Li DL, Li LC, Xu L, 264 Martınez-Sanchez N, Seoane-Collazo P, Contreras C, Zhao P & Gao J (2015) Inositol-requiring protein 1 - Varela L, Villarroya J, Rial-Pensado E, BuqueX, X-box-binding protein 1 pathway promotes epithelial- Aurrekoetxea I, Delgado TC, Vazquez-Mart ınez R mesenchymal transition via mediating snail expression et al. (2017) Hypothalamic AMPK-ER Stress-JNK1 in pulmonary fibrosis. Int J Biochem Cell Biol 65, 230– axis mediates the central actions of thyroid 238. hormones on energy balance. Cell Metab 26, 255 Minchenko DO, Kharkova AP, Halkin OV, 212–229.e12. Karbovskyi LL & Minchenko OH (2016) Effect of 265 Zughaier SM, Stauffer BB & McCarty NA (2014) hypoxia on the expression of genes encoding insulin- Inflammation and ER stress downregulate BDH2 like growth factors and some related proteins in u87 expression and dysregulate intracellular iron in glioma cells without IRE1 function. Endocr Regul 50, macrophages. J Immunol Res 2014, 140728. 43–54. 266 Mori K (2009) Signalling pathways in the unfolded 256 Li H, Chen X, Gao Y, Wu J, Zeng F & Song F protein response: development from yeast to (2015) XBP1 induces snail expression to promote mammals. J Biochem 146, 743–750. epithelial- to-mesenchymal transition and invasion of 267 Michalak M & Gye MC (2015) Endoplasmic breast cancer cells. Cell Signal 27,82–89. reticulum stress in periimplantation embryos. Clin Exp 257 Jin C, Jin Z, Chen N-Z, Lu M, Liu C-B, Hu W-L & Reprod Med 42,1–7. Zheng C-G (2016) Activation of IRE1a-XBP1 268 Li J, Chen Z, Gao LY, Colorni A, Ucko M, Fang S pathway induces cell proliferation and invasion in & Du SJ (2015) A transgenic zebrafish model for colorectal carcinoma. Biochem Biophys Res Commun monitoring xbp1 splicing and endoplasmic reticulum 470,75–81. stress in vivo. Mech Dev 137,33–44. 258 Feng YX, Sokol ES, Del Vecchio CA, Sanduja S, 269 Ishikawa T, Kashima M, Nagano AJ, Ishikawa- Claessen JHL, Proia TA, Jin DX, Reinhardt F, Ploegh Fujiwara T, Kamei Y, Todo T & Mori K (2017) HL, Wang Q et al. (2014) Epithelial-to-mesenchymal Unfolded protein response transducer IRE1-mediated transition activates PERK-eIF2a and sensitizes cells to signaling independent of XBP1 mRNA splicing is not endoplasmic reticulum stress. Cancer Discov 4, 702– required for growth and development of medaka fish. 715. Elife 6,1–29. 259 Moon SY, Kim HS, Nho KW, Jang YJ & Lee SK 270 Richardson CE, Kinkel S & Kim DH (2011) (2014) Endoplasmic reticulum stress induces epithelial- Physiological IRE-1-XBP-1 and PEK-1 signaling in mesenchymal transition through autophagy via Caenorhabditis elegans larval development and activation of c-Src kinase. Nephron Exp Nephrol 126, immunity. PLoS Genet 7,1–10. 127–140. 271 Shi W, Xu G, Wang C, Sperber SM, Chen Y, Zhou 260 Greenwood M, Bordieri L, Greenwood MP, Rosso Q, Deng Y & Zhao H (2015) Heat shock 70-kDa Melo M, Colombari DS, Colombari E, Paton JFR & protein 5 (Hspa5) is essential for pronephros

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 273 Federation of European Biochemical Societies.

46 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

formation by mediating retinoic acid signaling. J Biol stress mediates the differentiation of fibroblasts. PLoS Chem 290, 577–589. ONE 10,1–11. 272 Luo S, Mao C, Lee B & Lee AS (2006) GRP78/BiP 284 Arensdorf AM, Diedrichs D & Rutkowski DT (2013) Is required for cell proliferation and protecting the Regulation of the transcriptome by ER stress: non- inner cell mass from apoptosis during early mouse canonical mechanisms and physiological consequences. embryonic development. Mol Cell Biol 26, Front Genet 4,1–16. 5688–5697. 285 Murakami T, Saito A, Hino S, Kondo S, Kanemoto S, 273 Zhang X, Szabo E, Michalak M & Opas M (2007) Chihara K, Sekiya H, Tsumagari K, Ochiai K, Endoplasmic reticulum stress during the embryonic Yoshinaga K et al. (2009) Signalling mediated by the development of the central nervous system in the endoplasmic reticulum stress transducer OASIS is mouse. Int J Dev Neurosci 25, 455–463. involved in bone formation. Nat Cell Biol 11, 1205–1211. 274 Kratochvılova K, Moran L, Pad’ourova S, Stejskal S, 286 Hamamura K & Yokota H (2007) Stress to TesarovaL, Simara P, Hampl A, KoutnaI& endoplasmic reticulum of mouse osteoblasts induces Vanhara P (2016) The role of the endoplasmic apoptosis and transcriptional activation for bone reticulum stress in stemness, pluripotency and remodeling. FEBS Lett 581, 1769–1774. development. Eur J Cell Biol 95, 115–123. 287 Cui M, Kanemoto S, Cui X, Kaneko M, Asada R, 275 Hao L, Vassena R, Wu G, Han Z, Cheng Y, Latham Matsuhisa K, Tanimoto K, Yoshimoto Y, Shukunami KE & Sapienza C (2009) The unfolded protein C & Imaizumi K (2015) OASIS modulates hypoxia response contributes to preimplantation mouse pathway activity to regulate bone angiogenesis. Sci embryo death in the DDK syndrome. Biol Reprod 80, Rep 5, 16455. 944–953. 288 Kondo S, Murakami T, Tatsumi K, Ogata M, 276 Yang Y, Cheung HH, Tu J, Miu KK & Chan WY Kanemoto S, Otori K, Iseki K, Wanaka A & (2016) New insights into the unfolded protein response Imaizumi K (2005) OASIS, a CREB/ATF-family in stem cells. Oncotarget 7, 54010–54027. member, modulates UPR signalling in astrocytes. Nat 277 Latham KE (2015) Endoplasmic reticulum stress Cell Biol 7, 186–194. signaling in mammalian oocytes and embryos: life in 289 Asada R, Saito A, Kawasaki N, Kanemoto S, balance. Int Rev Cell Mol Biol 316, 227–265. Iwamoto H, Oki M, Miyagi H, Izumi S & Imaizumi 278 Gao Y, Sartori DJ, Li C, Yu Q-C, Kushner JA, K (2012) The endoplasmic reticulum stress transducer Simon MC & Diehl JA (2012) PERK is required in OASIS is involved in the terminal differentiation of the adult pancreas and is essential for goblet cells in the large intestine. J Biol Chem 287, maintenance of glucose homeostasis. Mol Cell Biol 8144–8153. 32, 5129–5139. 290 Saito A (2014) Physiological functions of endoplasmic 279 Bettigole SE, Lis R, Adoro S, Lee AH, Spencer LA & reticulum stress transducer OASIS in central nervous Weller PFGL (2015) The transcription factor XBP1 is system. Anat Sci Int 89,11–20. selectively required for eosinophil differentiation. Nat 291 van der Harg JM, van Heest JC, Bangel FN, Patiwael Immunol 16, 829–837. S, van Weering JRT & Scheper W (2017) The UPR 280 Leung A, Gregory NS, Allen L-AH & Sluka KA reduces glucose metabolism via IRE1 signaling. (2017) Regular physical activity prevents chronic pain Biochim Biophys Acta 1864, 655–665. by altering muscle macrophage phenotype and 292 Topisirovic I & Sonenberg N (2011) mRNA increasing IL-10 in mice. Pain 157,70–79. translation and energy metabolism in cancer: the role 281 Todd DJ, McHeyzer-Williams LJ, Kowal C, Lee AH, of the MAPK and mTORC1 Pathways. Cold Spring Volpe BT, Diamond B, McHeyzer-Williams MG & Harb Symp Quant Biol 76, 355–367. Glimcher LH (2009) XBP1 governs late events in 293 Lowe CE, Dennis RJ, Obi U, O’Rahilly S & plasma cell differentiation and is not required for Rochford JJ (2012) Investigating the involvement of antigen-specific memory B cell development. J Exp the ATF6a pathway of the unfolded protein Med 206, 2151–2159. response in adipogenesis. Int J Obes (Lond) 36, 282 van Galen P, Kreso A, Mbong N, Kent DG, 1248–1251. Fitzmaurice T, Chambers JE, Xie S, Laurenti E, 294 Lipson KL, Ghosh R & Urano F (2008) The role of Hermans K, Eppert K et al. (2014) The unfolded IRE1a in the degradation of insulin mRNA in protein response governs integrity of the pancreatic b-cells. PLoS ONE 3,1–7. haematopoietic stem-cell pool during stress. Nature 295 Zhang N, Yang X, Yuan F, Zhang L, Wang Y, Wang 510, 268–272. L, Mao Z, Luo J, Zhang H, Zhu WG et al. (2018) 283 Matsuzaki S, Hiratsuka T, Taniguchi M, Shingaki K, Increased amino acid uptake supports autophagy- Kubo T, Kiya K, Fujiwara T, Kanazawa S, deficient cell survival upon glutamine deprivation. Cell Kanematsu R, Maeda T et al. (2015) Physiological ER Rep 10, 3006–3020.

274 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

47 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

296 Rubio-Patino~ C, Bossowski JP, De Donatis GM, 306 Cross BC, Bond PJ, Sadowski PG, Jha BK, Zak J, Mondragon L, Villa E, Aira LE, Chiche J, Mhaidly Goodman JM, Silverman RH, Neubert TA, Baxendale R, Lebeaupin C, Marchetti S et al. (2018) Low-protein IR, Ron D et al. (2012) The molecular basis for diet induces IRE1a-dependent anticancer selective inhibition of unconventional mRNA splicing immunosurveillance. Cell Metab 27, 828–842. by an IRE1-binding small molecule. Proc Natl Acad 297 Xue Z, He Y, Ye K, Gu Z, Mao Y & Qi L (2011) A Sci USA 109, E869–E878. conserved structural determinant located at the 307 Mimura N, Fulciniti M, Gorgun G, Tai Y-T, Cirstea interdomain region of mammalian inositol-requiring D, Santo L, Hu Y, Fabre C, Minami J, Ohguchi H enzyme 1a. J Biol Chem 286, 30859–30866. et al. (2012) Blockade of XBP1 splicing by inhibition 298 Waller DD, Jansen G, Golizeh M, Martel-Lorion C, of IRE1 is a promising therapeutic option in multiple Dejgaard K, Shiao TC, Mancuso J, Tsantrizos YS, myeloma. Blood 119, 5772–5781. Roy R, Sebag M et al. (2016) A covalent cysteine- 308 Papandreou I, Denko NC, Olson M, Van Melckebeke targeting kinase inhibitor of Ire1 permits allosteric H, Lust S, Tam A, Solow-Cordero DE, Bouley DM, control of endoribonuclease activity. ChemBioChem Offner F, Niwa M et al. (2011) Identification of an 17, 843–851. Ire1alpha endonuclease specific inhibitor with 299 Wiseman RL, Zhang Y, Lee KPK, Harding HP, cytotoxic activity against human multiple myeloma. Haynes CM, Price J, Sicheri F & Ron D (2010) Blood 117, 1311–1314. Flavonol activation defines an unanticipated ligand- 309 Ri M, Tashiro E, Oikawa D, Shinjo S, Tokuda M, binding site in the kinase-RNase domain of IRE1. Yokouchi Y, Narita T, Masaki A, Ito A, Ding J et al. Mol Cell 38, 291–304. (2012) Identification of Toyocamycin, an agent 300 Rong J, Pass I, Diaz PW, Ngo TA, Sauer M, cytotoxic for multiple myeloma cells, as a potent Magnuson G, Zeng F-Y, Hassig CA, Jackson MR, inhibitor of ER stress-induced XBP1 mRNA splicing. Cosford NDP et al. (2015) Cell-based high-throughput Blood Cancer J 2, e79. luciferase reporter gene assays for identifying and 310 Tomasio SM, Harding HP, Ron D, Cross BCS & profiling chemical modulators of endoplasmic Bond PJ (2013) Selective inhibition of the unfolded reticulum signaling protein, IRE1. J Biomol Screen 20, protein response: targeting catalytic sites for Schiff 1232–1245. base modification. Mol BioSyst 9, 2408–2416. 301 Feldman HC, Tong M, Wang L, Meza-Acevedo R, 311 Jiang D, Lynch C, Medeiros BC, Liedtke M, Bam R, Gobillot TA, Lebedev I, Gliedt MJ, Hari SB, Mitra Tam AB, Yang Z, Alagappan M, Abidi P, Le Q-T AK, Backes BJ et al. (2016) Structural and functional et al. (2016) Identification of Doxorubicin as an analysis of the allosteric inhibition of IRE1a with Inhibitor of the IRE1a-XBP1 Axis of the Unfolded ATP-competitive ligands. ACS Chem Biol 11, 2195– Protein Response. Sci Rep 6, 33353. 2205. 312 Park BK, Boobis A, Clarke S, Goldring CEP, Jones 302 Harrington PE, Biswas K, Malwitz D, Tasker AS, D, Kenna JG, Lambert C, Laverty HG, Naisbitt DJ, Mohr C, Andrews KL, Dellamaggiore K, Kendall R, Nelson S et al. (2011) Managing the challenge of Beckmann H, Jaeckel P et al. (2015) Unfolded protein chemically reactive metabolites in drug development. response in cancer: IRE1a inhibition by selective Nat Rev Drug Discov 10, 292–306. kinase ligands does not impair tumor cell viability. 313 Wilson AJ (2009) Inhibition of protein–protein ACS Med Chem Lett 6,68–72. interactions using designed molecules. Chem Soc Rev 303 Wang L, Perera BGK, Hari SB, Bhhatarai B, Backes 38, 3289. BJ, Seeliger MA, Schurer€ SC, Oakes SA, Papa FR & 314 Atkins C, Liu Q, Minthorn E, Zhang S-Y, Figueroa Maly DJ (2012) Divergent allosteric control of the DJ, Moss K, Stanley TB, Sanders B, Goetz A, Gaul IRE1a endoribonuclease using kinase inhibitors. Nat N et al. (2013) Characterization of a novel PERK Chem Biol 8, 982–989. kinase inhibitor with antitumor and antiangiogenic 304 Jha BK, Polyakova I, Kessler P, Dong B, Dickerman activity. Cancer Res 73, 1993–2002. B, Sen GC & Silverman RH (2011) Inhibition of 315 Axten JM, Medina JR, Feng Y, Shu A, Romeril SP, RNase L and RNA-dependent protein kinase (PKR) Grant SW, Li WH, Heerding DA, Minthorn E, by sunitinib impairs antiviral innate immunity. J Biol Mencken T et al. (2012) Discovery of 7-methyl-5-(1- Chem 286, 26319–26326. {[3-(trifluoromethyl)phenyl]acetyl}-2,3-dihydro-1H- 305 Volkmann K, Lucas JL, Vuga D, Wang X, Brumm D, indol-5-yl)-7H-pyrrolo[2,3-d]pyrimidin-4-amine Stiles C, Kriebel D, Der-Sarkissian A, Krishnan K, (GSK2606414), a potent and selective first-in-class Schweitzer C et al. (2011) Potent and selective inhibitor of protein kinase R (PKR)-like endoplasmic inhibitors of the inositol-requiring enzyme 1 reticulum kinase (PERK). J Med Chem 55, 7193– endoribonuclease. J Biol Chem 286, 12743–12755. 7207.

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 275 Federation of European Biochemical Societies.

48 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

316 Moreno JA, Halliday M, Molloy C, Radford H, (2012) Discovery of an orally active small-molecule Verity N, Axten JM, Ortori CA, Willis AE, Fischer irreversible inhibitor of protein disulfide isomerase for PM, Barrett DA et al. (2013) Oral treatment targeting ovarian cancer treatment. Proc Natl Acad Sci 109, the unfolded protein response prevents 16348–16353. neurodegeneration and clinical disease in prion- 327 Banerjee R, Pace NJ, Brown DR & Weerapana E infected mice. Sci Transl Med 138, 206ra138. (2013) 1,3,5-Triazine as a modular scaffold for 317 Harding HP, Zeng H, Zhang Y, Jungries R, Chung covalent inhibitors with streamlined target P, Plesken H, Sabatini DD & Ron D (2001) identification. J Am Chem Soc 135, 2497–2500. Diabetes mellitus and exocrine pancreatic 328 Chevet E, Hetz C & Samali A (2015) Endoplasmic dysfunction in perk-/- mice reveals a role for reticulum stress-activated cell reprogramming in translational control in secretory cell survival. Mol oncogenesis. Cancer Discov 5, 586–597. Cell 7, 1153–1163. 329 Okada T, Haze K, Nadanaka S, Yoshida H, Seidah 318 Rojas-Rivera D, Delvaeye T, Roelandt R, Nerinckx NG, Hirano Y, Sato R, Negishi M & Mori K (2003) W, Augustyns K, Vandenabeele P & Bertrand MJ A serine protease inhibitor prevents endoplasmic (2017) When PERK inhibitors turn out to be new reticulum stress-induced cleavage but not transport of potent RIPK1 inhibitors: critical issues on the the membrane-bound transcription factor ATF6. J specificity and use of GSK2606414 and GSK2656157. Biol Chem 278, 31024–31032. Cell Death Differ 24, 1100–1110. 330 Doultsinos D, Avril T, Lhomond S, Dejeans N, 319 Sidrauski C, Tsai JC, Kampmann M, Hearn BR, Guedat P & Chevet E (2017) Control of the unfolded Vedantham P, Jaishankar P, Sokabe M, Mendez AS, protein response in health and disease. SLAS Discov Newton BW, Tang EL et al. (2015) Pharmacological 22, 787–800. dimerization and activation of the exchange factor 331 Delpino A & Castelli M (2002) The 78 kDa glucose- eIF2B antagonizes the integrated stress response. Elife regulated protein (GRP78/BIP) is expressed on the cell 2015, e07314. membrane, is released into cell culture medium and is 320 Sekine Y, Zyryanova A, Crespillo-Casado A, Fischer also present in human peripheral circulation. Biosci PM, Harding HP & Ron D (2015) Stress responses. Rep 22, 407–420. Mutations in a translation initiation factor identify the 332 Chessler SD & Byers PH (1993) BiP binds type I target of a memory-enhancing compound. Science 348, procollagen pro alpha chains with mutations in the 1027–1030. carboxyl-terminal propeptide synthesized by cells from 321 Halliday M, Radford H, Sekine Y, Moreno J, Verity patients with osteogenesis imperfecta. J Biol Chem N, Le Quesne J, Ortori CA, Barrett DA, Fromont C, 268, 18226–18233. Fischer PM et al. (2015) Partial restoration of protein 333 Booth L, Roberts JL, Cash DR, Tavallai S, Jean S, synthesis rates by the small molecule ISRIB prevents Fidanza A, Cruz-Luna T, Siembiba P, Cycon KA, neurodegeneration without pancreatic toxicity. Cell Cornelissen CN et al. (2015) GRP78/BiP/HSPA5/ Death Dis 6, e1672. DNA K is a universal therapeutic target for human 322 Sidrauski C, McGeachy AM, Ingolia NT & Walter P disease. J Cell Physiol 230, 1661–1676. (2015) The small molecule ISRIB reverses the effects 334 Zhu H, Chen X, Chen B, Chen B, Song W, Sun D & of eIF2a phosphorylation on translation and stress Zhao Y (2014) Activating transcription factor 4 granule assembly. Elife 2015,1–16. promotes esophageal squamous cell carcinoma 323 Gallagher CM, Garri C, Cain EL, Ang KKH, Wilson invasion and metastasis in mice and is associated with CG, Chen S, Hearn BR, Jaishankar P, Aranda-Diaz poor prognosis in human patients. PLoS ONE 9,1– A, Arkin MR et al. (2016) Ceapins are a new class of 11. unfolded protein response inhibitors, selectively 335 Yang J, Cheng D, Zhou S, Zhu B, Hu T & Yang Q targeting the ATF6a branch. Elife 5,1–33. (2015) Overexpression of X-Box binding protein 1 324 Gallagher CM & Walter P (2016) Ceapins inhibit (XBP1) correlates to poor prognosis and up-regulation ATF6a signaling by selectively preventing transport of of PI3K/mTOR in human osteosarcoma. Int J Mol ATF6a to the Golgi apparatus during ER stress. Elife Sci 16, 28635–28646. 5, e11880. 336 Bujisic B, De Gassart A, Tallant R, Demaria O, 325 Bu L, Yu H, Fan L, Li X, Wang F, Liu J, Zhong F, Zaffalon L, Chelbi S, Gilliet M, Bertoni F & Zhang C, Wei W, Wang H et al. (2017) Melatonin, a Martinon F (2017) Impairment of both IRE1 novel selective ATF-6 inhibitor, induces human expression and XBP1 activation is a hallmark of GCB hepatoma cell apoptosis through COX-2 DLBCL and contributes to tumor growth. Blood 129, downregulation. World J Gastroenterol 23, 986–998. 2420–2428. 326 Xu S, Butkevich AN, Yamada R, Zhou Y, Debnath 337 Kaser A, Lee A-H, Franke A, Glickman JN, Zeissig B, Duncan R, Zandi E, Petasis NA & Neamati N S, Tilg H, Nieuwenhuis EES, Higgins DE, Schreiber S,

276 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

49 A. Almanza et al. Compendium of endoplasmic reticulum stress signaling

Glimcher LH et al. (2008) XBP1 links ER stress to responsiveness characteristic of asthma. J Immunol intestinal inflammation and confers genetic risk for 192, 3475–3487. human inflammatory bowel disease. Cell 134, 348 Graner MW, Lillehei KO & Katsanis E (2015) 743–756. Endoplasmic reticulum chaperones and their roles in 338 Atkin JD, Farg MA, Walker AK, McLean C, Tomas the immunogenicity of cancer vaccines. Front Oncol 4, D & Horne MK (2008) Endoplasmic reticulum stress 1–12. and induction of the unfolded protein response in 349 Graner MW, Zeng Y, Feng H & Katsanis E (2003) human sporadic amyotrophic lateral sclerosis. Tumor-derived chaperone-rich cell lysates are effective Neurobiol Dis 30, 400–407. therapeutic vaccines against a variety of cancers. 339 Nardo G, Pozzi S, Pignataro M, Lauranzano E, Cancer Immunol Immunother 52, 226–234. Spano G, Garbelli S, Mantovani S, Marinou K, 350 Rapp UK & Kaufmann SH (2004) DNA vaccination Papetti L, Monteforte M et al. (2011) Amyotrophic with gp96-peptide fusion proteins induces protection lateral sclerosis multiprotein biomarkers in peripheral against an intracellular bacterial pathogen. Int blood mononuclear cells. PLoS ONE 6, e25545. Immunol 16, 597–605. 340 Kim Y, Lee H, Manson SR, Lindahl M, Evans B, 351 Qian J, Hong S, Wang S, Zhang L, Sun L, Wang M, Miner JH, Urano F & Chen YM (2016) Yang J & Kwak LW (2009) Myeloma cell line – Mesencephalic astrocyte-derived neurotrophic factor derived, pooled heat shock proteins as a universal as a urine biomarker for endoplasmic reticulum stress- vaccine for immunotherapy of multiple myeloma. related kidney diseases. J Am Soc Nephrol 27, 2974– Blood 114, 3880–3890. 2982. 352 Argon Y & Simen BB (1999) GRP94, an ER 341 Mami I, Bouvier N, El Karoui K, Gallazzini M, chaperone with protein and peptide binding Rabant M, Laurent-Puig P, Li S, Tharaux PL, Beaune properties. Semin Cell Dev Biol 10, 495–505. P, Thervet E et al. (2016) Angiogenin mediates cell- 353 Liu D, Liu X, Zhou T, Yao W, Zhao J, Zheng Z, autonomous translational control under endoplasmic Jiang W, Wang F, Aikhionbare FO, Hill DL et al. reticulum stress and attenuates kidney injury. JAm (2015) IRE1-RACK1 axis orchestrates ER stress Soc Nephrol 27, 863–876. preconditioning-elicited cytoprotection from ischemia/ 342 Dadey DYA, Kapoor V, Hoye K, Khudanyan A, reperfusion injury in liver. J Mol Cell Biol 8, 144–156. Collins A, Thotala D & Hallahan DE (2016) Antibody 354 Bailly-Maitre B, Fondevila C, Kaldas F, Droin N, targeting GRP78 enhances the efficacy of radiation Luciano F, Ricci J-E, Croxton R, Krajewska M, therapy in human glioblastoma and non-small-cell Zapata JM, Kupiec-Weglinski JW et al. (2006) lung cancer cell lines and tumor models. Clin Cancer Cytoprotective gene bi-1 is required for intrinsic Res 23, 2556–2564. protection from endoplasmic reticulum stress and 343 Lin YG, Shen J, Yoo E, Liu R, Yen HY, Mehta A, ischemia-reperfusion injury. Proc Natl Acad Sci USA Rajaei A, Yang W, Mhawech-Fauceglia P, DeMayo FJ 103, 2809–2814. et al. (2015) Targeting the glucose-regulated protein-78 355 Bi X, Zhang G, Wang X, Nguyen C, May HI, Li X, abrogates Pten-null driven AKT activation and Al-Hashimi AA, Austin RC, Gillette TG, Fu G et al. endometrioid tumorigenesis. Oncogene 34, 5418–5426. (2018) Endoplasmic reticulum chaperone GRP78 344 Chen YG, Ashok BT, Liu X, Garikapaty VPS, protects heart from ischemia/reperfusion injury Mittelman A & Tiwari RK (2003) Induction of heat through Akt activation. Circ Res 122, 1545–1554. shock protein gp96 by immune cytokines. Cell Stress 356 Martindale JJ, Fernandez R, Thuerauf D, Whittaker Chaperones 8, 242–248. R, Gude N, Sussman MA & Glembotski CC (2006) 345 Zhang K, Peng Z, Huang X, Qiao Z, Wang X, Wang Endoplasmic reticulum stress gene induction and N, Xi H, Cui J, Gao Y, Huang X et al. (2017) Phase protection from ischemia/reperfusion injury in the II trial of adjuvant immunotherapy with autologous hearts of transgenic mice with a tamoxifen-regulated tumor-derived Gp96 vaccination in patients with form of ATF6. Circ Res 98, 1186–1193. gastric cancer. J Cancer 8, 1826–1832. 357 Peralta C & Brenner C (2011) Endoplasmic reticulum 346 Tian S, Liu Z, Donahue C, Falo LD Jr & You Z stress inhibition enhances liver tolerance to ischemia/ (2012) Genetic targeting of the active transcription reperfusion. Curr Med Chem 18, 2016–2024. factor XBP1s to dendritic cells potentiates vaccine- 358 Folch-Puy E, Panisello A, Oliva J, Lopez A, Benıtez induced prophylactic and therapeutic antitumor CC, Adam R & Rosello-Catafau J (2016) Relevance of immunity. Mol Ther 20, 432–442. endoplasmic reticulum stress cell signaling in liver cold 347 Miller M, Rosenthal P, Beppu A, James L, Hoffman ischemia reperfusion injury. Int J Mol Sci 17,1–12. HM, Tam AB, Taylor A, Mcgeough MD, Pena CA, 359 Kwon SK, Ahn M, Song HJ, Kang SK, Jung SB, Niwa M et al. (2014) ORMDL3 transgenic mice have Harsha N, Jee S, Moon JY, Suh KS, Do LS et al. increased airway remodeling and airway (2015) Nafamostat mesilate attenuates transient focal

The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of 277 Federation of European Biochemical Societies.

50 Compendium of endoplasmic reticulum stress signaling A. Almanza et al.

ischemia/reperfusion-induced brain injury via the amyotrophic lateral sclerosis. Neurobiol Dis 71, 317– inhibition of endoplasmic reticulum stress. Brain Res 324. 1627,12–20. 369 Das I, Krzyzosiak A, Schneider K, Wrabetz L, 360 Zhang H, Yue Y, Sun T, Wu X & Xiong S (2017) D’Antonio M, Barry N & Sigurdardottir ABA (2015) Transmissible endoplasmic reticulum stress from Preventing proteostasis diseases by selective inhibition myocardiocytes to macrophages is pivotal for the of a phosphatase regulatory subunit. Science 348, pathogenesis of CVB3-induced viral myocarditis. Sci 239–242. Rep 7, 42162. 370 Lee A-H, Iwakoshi NN & Glimcher LH (2003) XBP-1 361 Rodvold JJ, Chiu KT, Hiramatsu N, Nussbacher JK, regulates a subset of endoplasmic reticulum resident Galimberti V, Mahadevan NR, Willert K, Lin JH & chaperone genes in the unfolded protein response. Mol Zanetti M (2017) Intercellular transmission of the Cell Biol 23, 7448–7459. unfolded protein response promotes survival and drug 371 Ghosh R, Wang L, Wang ES, Perera BGK, Igbaria resistance in cancer cells. Sci Signal 10, eaah7177. A, Morita S, Prado K, Thamsen M, Caswell D, https://doi.org/10.1126/scisignal.aah7177 Macias H et al. (2014) Allosteric inhibition of the 362 Brownlie RJ, Myers LK, Wooley PH, Corrigall VM, IRE1a RNase preserves cell viability and function Bodman-Smith MD, Panayi GS & Thompson SJ during endoplasmic reticulum stress. Cell 158, (2006) Treatment of murine collagen-induced arthritis 534–548. by the stress protein BiP Via interleukin-4 – producing 372 Kawamura T, Tashiro E, Shindo K & Imoto M regulatory t cells a novel function for an ancient (2008) SAR study of a novel triene-ansamycin group protein. Arthritis Rheumatol 54, 854–863. compound, quinotrierixin, and related compounds, as 363 Kirkham B, Chaabo K, Hall C, Garrood T, Mant T, inhibitors of ER stress-induced XBP1 activation II. Allen E, Vincent A, Vasconcelos JC, Prevost AT, Structure elucidation. J Antibiot (Tokyo) 61, 312–317. Panayi GS et al. (2016) Safety and patient response as 373 Chen D, Landis-Piwowar KR, Chen MS & Dou QP indicated by biomarker changes to binding (2007) Inhibition of proteasome activity by the dietary immunoglobulin protein in the phase I/IIA RAGULA flavonoid apigenin is associated with growth inhibition clinical trial in rheumatoid arthritis. Rheumatology in cultured breast cancer cells and xenografts. Breast (United Kingdom) 55, 1993–2000. Cancer Res 9, R80. 364 Axten JM, Romeril SP, Shu A, Ralph J, Medina JR, 374 Zhu M, Rajamani S, Kaylor J, Han S, Zhou F & Feng Y, Li WHH, Grant SW, Heerding DA, Fink AL (2004) The flavonoid baicalein inhibits Minthorn E et al. (2013) Discovery of GSK2656157: fibrillation of alpha-synuclein and disaggregates an optimized perk inhibitor selected for preclinical existing fibrils. J Biol Chem 279, 26846–26857. development. ACS Med Chem Lett 4, 964–968. 375 Kim D-S, Ha K-C, Kwon D-Y, Kim M-S, Kim H-R, 365 Boyce M, Bryant KF, Jousse C, Long K, Harding Chae S-W & Chae H-J (2008) Kaempferol protects HP, Scheuner D, Kaufman RJ, Ma D, Coen DM, ischemia/reperfusion-induced cardiac damage through Ron D et al. (2005) A selective inhibitor of eIF2alpha the regulation of endoplasmic reticulum stress. dephosphorylation protects cells from ER stress. Immunopharmacol Immunotoxicol 30, 257–270. Science 307, 935–939. 376 Plate L, Cooley CB, Chen JJ, Paxman RJ, Gallagher 366 Saxena S, Cabuy E & Caroni P (2009) A role for CM, Madoux F, Genereux JC, Dobbs W, Garza D, motoneuron subtype-selective ER stress in disease Spicer TP et al. (2016) Small molecule proteostasis manifestations of FALS mice. Nat Neurosci 12, 627– regulators that reprogram the ER to reduce 636. extracellular protein aggregation. Elife 5, e15550. 367 Colla E, Coune P, Liu Y, Pletnikova O, Troncoso JC, 377 Higa A, Taouji S, Lhomond S, Jensen D, Fernandez- Iwatsubo T, Schneider BL & Lee MK (2012) Zapico ME, Simpson JC, Pasquet JM, Schekman R & Endoplasmic reticulum stress is important for the Chevet E (2014) Endoplasmic reticulum stress- manifestations of a-synucleinopathy in vivo. J Neurosci activated transcription factor atf6a requires the 32, 3306–3320. disulfide isomerase PDIA5 to modulate 368 Wang L, Popko B, Tixier E & Roos RP (2014) chemoresistance. Mol Cell Biol 34, 1839–1849. Guanabenz, which enhances the unfolded protein response, ameliorates mutant SOD1-induced

278 The FEBS Journal 286 (2019) 241–278 ª 2018 The Authors. The FEBS Journal published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

51 Chapter 1: Conclusions and Contributions

In this first chapter we have documented ER knowledge which led to its emerging roles in physiology and disease, encompassing interorgan communication and exploitation opportunities in bioengineering and health, outcomes of which include cost effective biologics, crop or livestock improvement and the provision of precision tools to guide therapeutic innovation and stratification in specific patient pools. We have looked at how studying ER stress signalling in different organisms is enabled by its conserved nature in metazoans and therefore how, in turn, research in various organisms is complemented by high throughput and strategies to produce and models, supplying us with data for pre-clinical scrutiny and subsequent clinical application. It is thus evident that understanding ER signalling and the UPR response, can enhance healthcare at a preventative, diagnostic and therapeutic level. In the next chapter we will be focusing further on the UPR and IRE1 and the ways these can be targeted.

The author of this thesis contributed to the publication of “Endoplasmic reticulum stress signalling – from basic mechanisms to clinical applications” the following: Original text for a share of main text attributed to “Chapter 1.5 Physiological ER signalling”, original text for the “Abstract” and “Concluding Remarks” as well as original text for the cover letter and review outline submitted to the FEBS Journal initiating correspondence with the Editor-in- charge. Artistically he provided Figures 2 and 33. He was responsible for the management of the group responsible for a third of the total manuscript corresponding to Chapters 1.4 and 1.5 exclusively and 1.3, 1.6 partially coordinating with the other project managers to provide project deliverables4. Finally he provided extensive writing quality control, editorial duties and coordinated efforts to produce a homogeneous manuscript of publishable quality5.

3 Special mention to Maria Livia Sassano for her input in shaping figure 3. 4 Special thanks to Dr E. Chevet and A. Papaioannou, C. Chintha who as overall project manager and group managers for chapters 1.1, 1.2, 1.3 and 1.6, 1.7 respectively, helped to bring this review to term. 5 Special thanks to Dr E. Chevet, Dr S. Healy and A. Papaioannou, C. Chintha for their contribution to shaping the final manuscript

52 Chapter 2

6Control of the Unfolded Protein Response in Health and Disease

6 The artwork for the cover of this issue of EMBO Molecular Medicine is owned by SAGE Publishing.

53 Chapter 2: Foreword

54 JBXJBJ XXX X10.1177/2472555217701685SLAS11000..1.1117777/27/2424747247257255555552522177777070016016855SSLLASLAS Discovery:covery:cooovovevery:vvee y AdvancinAdvanciAdvaAdvancingg Life Sciences R&DDoultsinosultsinosltt ett alaal. research-article7016852017 68585

Review

SLAS Discovery 2017, Vol. 22(7) 787 –800 Control of the Unfolded Protein Response © 2017 Society for Laboratory Automation and Screening DOI:https://doi.org/10.1177/2472555217701685 10.1177/2472555217701685 in Health and Disease journals.sagepub.com/home/jbx

Dimitrios Doultsinos1,2, Tony Avril1,2, Stéphanie Lhomond3, Nicolas Dejeans3, Philippe Guédat4, and Eric Chevet1,2,3

Abstract The unfolded protein response (UPR) is an integrated, adaptive biochemical process that is inextricably linked with cell homeostasis and paramount to maintenance of normal physiological function. Prolonged accumulation of improperly folded proteins in the endoplasmic reticulum (ER) leads to stress. This is the driving stimulus behind the UPR. As such, prolonged ER stress can push the UPR past beneficial functions such as reduced protein production and increased folding and clearance to apoptotic signaling. The UPR is thus contributory to the commencement, maintenance, and exacerbation of a multitude of disease states, making it an attractive global target to tackle conditions sorely in need of novel therapeutic intervention. The accumulation of information of screening tools, readily available therapies, and potential pathways to drug development is the cornerstone of informed clinical research and clinical trial design. Here, we review the UPR’s involvement in health and disease and, beyond providing an in-depth description of the molecules found to target the three UPR arms, we compile all the tools available to screen for and develop novel therapeutic agents that modulate the UPR with the scope of future disease intervention.

Keywords endoplasmic reticulum, stress, UPR, pharmacology, screening

Introduction such, protein homeostasis (proteostasis) in the ER is ensured by the coordinated action of chaperoning, folding, quality The endoplasmic reticulum (ER) is a large cellular organ- control (QC), and degradation mechanisms.5 Newly synthe- elle that can account for more than 50% of total membranes sized proteins enter the ER, where they face a folding prone in eukaryotic cells. It was one of the last cellular organelles 1 environment comprising chaperones (e.g., glucose regu- to be discovered in 1945 by Porter, Claude, and Thompson lated proteins such as GRP78/BiP, GRP94, GRP170) and and is one of the most structurally complex and evolution- enzymatic complexes involved in posttranslational modifi- arily diverse components of cellular function and mainte- cation such as N-linked glycosylation (e.g., oligosaccharyl nance. It is subcategorized into smooth ER, the specialized transferase complex) or disulphide bond formation (e.g., myocyte sarcoplasmic reticulum, and rough ER (RER). The protein disulphide isomerases).6 These newly synthesized RER is distinguished by ER membrane–bound ribosomes proteins undergo a folding process, which, if successful, that play a significant role in secretory and transmembrane 2 allows them to exit the ER and traffic toward their final protein biogenesis. The smooth ER is a site of sterol and destination through the secretory pathway.7 If folding fails, lipid production, thereby majorly contributing phospholip-

ids to mitochondrial and peroxisomal membranes as well as 1 the Golgi apparatus, secretory vesicles, and, of course, the Inserm U1242, Chemistry, Oncogenesis, Stress & Signaling, University 3 of Rennes 1, Rennes, France ER itself. Moreover, it is the largest depository of intracel- 2Centre de Lutte contre le Cancer Eugène Marquis, Rennes, France lular calcium, paramount in a vast host of processes such as 3BMYscreen, Bergonié Cancer Institute, Bordeaux, France muscle fiber contraction, mitochondrial function, neu- 4Inflectis Bioscience, Nantes, France rotransmitter release, and the regulation of the secretory Received Dec 16, 2016, and in revised form Mar 2, 2017, Accepted for pathway through signaling the release of stored proteins publication Mar 6, 2017. from secretory vesicles.4 Collectively, the ER integrates all of these different properties to fulfill its main task that lies Corresponding Author: in the productive folding of secretory and transmembrane Eric Chevet, Inserm U1242, Chemistry, Oncogenesis, Stress & Signaling, Centre de Lutte contre le Cancer Eugène Marquis, Avenue de la bataille proteins. This highly active processing must be accompa- Flandres Dunkerque, F-35000 Rennes, France. nied by a finely tuned regulation of ER homeostasis. As Email: [email protected]

55 788 SLAS Discovery 22(7)

Figure 1. The three endoplasmic reticulum (ER) stress sensors (PERK [red], IRE1 [blue], ATF6 [green]) initially activate signaling events that increase protein-folding capacity and reduce protein load on the ER. These transcriptional and translational outputs tend to reestablish protein- folding homeostasis in the ER and promote cell survival.14,24 improperly folded proteins are retained in the ER through The UPR: Canonical and Noncanonical 8 QC mechanisms. For instance, a specific QC machinery Signaling Pathways has been identified for N-linked glycoproteins that includes the lectins calnexin and calreticulin, which bind monoglu- The canonical core of the UPR is composed of three major 16–18 cosylated glycans, and the UDP-glucose glucosyl transfer- transmembrane stress sensors; PERK, IRE-1, and ATF6. ases, whose role is to re-monoglucosylate improperly folded In their inactive state, all three sensors are bound to the pro- 19–21 glycoproteins to provide them with another chance of tein BiP, otherwise known as GRP78 or HSPA5. BiP is acquiring a proper conformation.9–11 Terminally misfolded an ER chaperone that is indispensable to both basal and per- proteins, identified as such as a result of several failed fold- turbed UPR as it displays antiapoptotic abilities, is heavily ing attempts, are then retrotranslocated in the cytosol to be involved in calcium homeostasis, and promotes QC by degraded via the proteasome. This process is called either targeting misfolded proteins for ERAD-mediated ER-associated degradation (ERAD).12 Under homeostatic degradation or promoting correct refolding and transloca- 22,23 conditions, ER folding, export, and degradation are not tion. Upon stress, BiP dissociates from the three sensors saturated and cope with the protein folding demand. of the UPR, allowing them to trigger prosurvival or, in the However, during certain physiological and pathological case of overwhelming stress, prodeath mechanisms. It is conditions, the protein misfolding burden overwhelms ER this function that has led to BiP to be considered a master folding and export capacity, thereby leading to an accumu- regulator of the UPR. The three sensors and associated sig- lation of improperly folded proteins in this compartment naling pathways can be viewed in Figure 1. and to a situation called ER stress. To cope with ER stress, the ER has evolved an adaptive pathway, called the IRE1 unfolded protein response (UPR), whose main function is to restore protein homeostasis in this compartment.13 If the IRE1 in humans exists in the form of two homologues: IRE1α stress is alleviated, the UPR is shut down and ER proteos- and IRE1β.25 IRE1α (referred to as IRE1 hereafter) is ubiqui- tasis is restored.14 In contrast, if adaptation fails and tously and abundantly expressed, whereas IRE1β is abundant homeostatic mechanisms are overcome, the UPR also trig- only in intestinal and pulmonary epithelial and mucosal epithe- gers signals that will engage and destine cells to die through lial tissues.26,27 Interestingly, IRE1β displays higher structural apoptosis.15 similarity to murine IRE1 than human IPE1D.28 IRE1 is a

56 Doultsinos et al. 789 transmembrane type 1 protein. Its cytosolic domain carries out domain. Upon ER stress, ATF6 is exported to the Golgi two enzymatic activities including serine/threonine kinase and apparatus, where it is cleaved on both sides of the mem- endoribonuclease activities.29 Recently, the activation profiles brane by the proteases S1P and S2P.44,45 This prompts the of IRE1 have revealed that the protein can form dimers and release of the ATF6 cytosolic domain (ATF6f), which then oligomers, which in turn confer selectivity in the output signals translocates to the nucleus to act as a transcription factor emitted from IRE1.30 This includes not only the nonconven- for genes whose products are involved in ERAD and qual- tional splicing of XBP1 mRNA (which was recently found to ity control (HERPUD1, SEL1L, OS9), folding (BiP, be the result of the concerted action of IRE1 RNase and the GRP94), and redox control (PDIA4, ERO1L).46 Beyond RTCB tRNA ligase)31 but also the regulated IRE1-dependent the role of BiP in ATF6 activation, the protein disulphide decay of RNA (RIDD, including mRNA32 and microRNA14). isomerase PDIA5 has also been shown to be instrumental The coordinated action of these IRE1-dependent mechanisms in this process.47 yields select biological outputs. Interestingly, upon ER stress, In recent years, other events were described to transduce the splicing of XBP1 mRNA has been shown to be transient information about the ER status independently of the three and the subject of inactivation mechanisms that depend in part UPR sensors, leading to their attribution to the noncanoni- on the protein disulphide isomerase PDIA6.33 In contrast, cal UPR. Those three different classes can be defined as RIDD activity, which constitutively exists (tonic RIDD) (1) regulation of calcium fluxes, (2) transmembrane tran- appears to increase with stress, thereby leading to an uncon- scription factors, and (3) atypical transcription mechanisms. trolled and death-promoting mechanism (prodeath RIDD).34 First, calcium release from the ER has been shown to repre- The precise mechanisms by which IRE1 operates XBP1 sent an atypical signaling event upon ER stress. Indeed, ino- mRNA splicing and RIDD remain unclear. sitol 3-phosphate–induced calcium release is increased during ER stress and is also involved in CHOP-mediated 48 PERK apoptotic signaling through ER oxidases as well as being linked with the binding to members of the apoptotic BCL-2 PERK is similarly activated to IRE1, and it phosphorylates family to modulate calcium release and modulate ER stress eukaryotic translation initiation factor 2α (eIF2α) that ulti- cell death.49 Therefore, dysregulation of calcium fluxes can mately inhibits protein translation and thus synthesis as a lead to the initiation or exacerbation of the maladaptive potent prosurvival mechanism; this is based on the fact that the UPR and hence lead to apoptotic cell death. Moreover, number of proteins entering the ER for processing is massively apoptotic regulators such as the Golgi anti-apoptotic protein reduced.35 In contrast to its induction of a generalized arrest in modulate intracellular calcium fluxes50 and caspases 4 and translation, eIF2α phosphorylation promotes the translation of 12 have been shown to be activated during ER stress in a the transcription factor ATF4.36 ATF4 then acts as either a tran- calcium-dependent manner in Parkinsonism and prion neu- scriptional repressor or a transcriptional activator and affects a rotoxic disorders.51,52 Beyond ATF6, a large family of trans- multitude of biological processes such as bone resorption membrane bZIP proteins has been discovered and plays (RANKL),37 amino acid metabolism (asparagine synthase),38 diverse functions in the ER homeostasis outside the scope neovascularization (VEGF),39 as well as controlling expression of the canonical UPR.53 This family comprises the tran- of genes involved in ER chaperone and foldase production.14 scription factors OASIS, BBF2H7, CREBH, AIbZIP, and CHOP and GADD34 further contribute to PERK-mediated Luman and has been implicated in diverse responses to cell cell fate control by mediating apoptosis or PERK inhibition, differentiation, maturation, and homeostasis in different tis- respectively.40 In a study by Novoa et al.,41 phosphorylated sue distributions.54 Each family member may play a distinct eIF2α was found to be greatly diminished in GADD34 overex- role in ER homeostasis, and it has been shown in an in vivo pressing cells because of the formation of a complex between model of cerebral ischemia that BB2H7 is predominantly GADD34 and the catalytic domain of protein phosphatase 1 present in the peri-infarction region, pointing to an up- (PP1c) that resulted in the dephosphorylation of eIF2α and regulation during a specific phase of the UPR.55 Another thus blocked attenuation of CHOP and interfered with ATF4 example is provided by CREBH, whose activation upon ER activation in a stress-dependent manner. As is the case with stress was shown to control iron regulation in the liver IRE1, PDIA6 controls the transition to the inactive form of the through the control of hepcidin expression.56 More recently, protein.42 In addition to eIF2α, PERK has been shown to target the AAA+ ATPase CDC48(yeast)/valosin-containing pro- Nrf2, thereby mediating the antioxidant responsive element tein (VCP)/p97(mammals), known to play a role in ERAD,57 response.43 was shown to also regulate ER stress-induced gene expression.58 This mechanism was shown to occur through the ATF6 p97/VCP-mediated degradation of RUVBL-2, another AAA+ ATPase acting as a repressor of XBP1s-dependent ATF6 is a type 2 transmembrane protein that resides in the transcription, thus indicating a dual role for p97/VCP in ER. It contains a transcription factor in its cytosolic protein degradation and transcriptional control.59

57 790 SLAS Discovery 22(7)

Control of ER Proteostasis in Diseases support this, administration of the eIF2α phosphorylation inducer salubrinal to SOD1 transgenic mice arrests disease The UPR has been extensively implicated in disease either progression, PERK haploinsufficiency exacerbates disease, as a cause or a consequence, playing roles in disease com- and XBP1 down-regulation stalls disease onset and mencement or exacerbation. These effects are mainly medi- 71 60 improves prognosis. Furthermore, PDI and GADD34 ated by prodeath and prosurvival properties of the UPR. In have been shown in in vivo models upon disease onset to certain clinical scenarios, prodeath signaling leads to degen- disseminate into ventral horn astrocytes and white matter eration and apoptosis, whereas in others, the same biochem- microglia.72 ALS is a complex spectrum disorder of unclear ical pathways lead to disease therapeutic resistance. origin. As a predominantly sporadic (>80%) condition diag- Amyotrophic lateral sclerosis (ALS; reviewed in ref. 61) nosed late in its progression, it is extremely difficult to dis- and glioblastoma multiforme (GBM; reviewed in ref. 62) cern a specific pathognomonic pattern and therefore decide are two examples of this diverse response in the central ner- whether UPR involvement is causative or consequential. In vous system. P97/VCP, involved in the noncanonical con- some cases, it may well be true that it is both, encompassing trol of the UPR, provides a genetic crossover between ALS a positive feedback loop of prodeath signals that are com- and GBM. It has been shown to be able to induce ALS with pounded by a multitude of exacerbating conditions that or without frontotemporal lobar degeneration accompanied arise as part of the ALS comorbidity spectrum (frontotem- by Paget disease, mimicking osteopathies and inclusion poral dementia, osteopathies, and gastrointestinal distur- body myositis by colocalizing in trademark ALS cytoplas- bances due to bulbar involvement, among others). Therefore, mic protein inclusions with established common ALS pro- 63,64 targeting the UPR at this postdiagnostic stage would play a teins FUS and TDP-43. Conversely, VCP has been disease-limiting role. By blocking the UPR prodeath sig- shown to regulate DNA-dependent protein kinase degrada- 65 nals that dominate ALS progression, the patient’s prognosis tion and hence affect radiation sensitivity of GBM cells. could be drastically improved. Despite this crossover, the hyperactivation of the perturbed Diabetes mellitus is a disease that affects almost 430 UPR in these diseases has contrasting effects, leading to million people worldwide.73 The development of diabetes autophagy and apoptosis of motor neurons in ALS but to depends on loss of pancreatic β cells, which are the cells chemoresistance, neoangiogenesis, and migration regula- responsible for producing insulin to regulate glucose tion in cancer cells in GBM. We look into the prodeath and homeostasis in the islets of Langerhans. To respond to such prosurvival effect of the UPR in different diseases in more high acute or chronic metabolic and proteostatic demand, β detail below. cells have an established ER that has been shown to be paramount in the homeostasis of β cells and a contributor in the Prodeath Signaling of the UPR in Diseases: development of diabetic symptoms when it fails.74 The Neurodegeneration and Diabetes knockout of IRE1 and XBP1 in β cells produces a distinct decline in cell proliferation and also affects proinsulin and Classical ALS is a predominantly adult-onset neurodegen- insulin synthesis and secretion/excretion.75 Similarly, erative genetic disorder characterized by rapidly progres- PERK inactivation displays diabetic symptoms of hyper- sive loss of both upper and lower motor neurons, with the glycaemia in in vivo models, and the UPR in general is trig- most common cause of death being respiratory failure fol- gered during the misfolding of preproinsulin, eventually 66 lowing diaphragmatic muscle failure. Studies in in vivo leading to accumulated and overwhelmed ER stress and β models of ALS have shown that pharmacologically modu- cell apoptosis.76 In the case of diabetes, targeting the UPR lating the UPR can enhance ALS pathogenesis or indeed could play both a disease-limiting and a preventative role. 67 replicate it. Superoxide dismutase (SOD1) mutations lead Diabetes is a small- and medium-vessel disease, and the to severe oxidative stress by affecting the conversion of UPR has been strongly linked with VEGF regulation and superoxide to water or hydrogen peroxide and the regula- neoangiogenesis.77 By targeting these pathways, the quality 68 tion of copper as a catalyst for SOD1. Furthermore, the of life of patients already suffering from diabetes could be trademarks of TDP43 and VCP mutation causing ALS are vastly improved as harnessing and limiting such processes cytoplasmic ubiquitinated inclusion protein aggregates and could have an effect on manifestations such as diabetic reti- autophagy. Oxidative stress, autophagy, and protein aggre- nopathy and coronary artery disease. gation are trademark triggers of ER stress and subsequently the UPR.69 Furthermore, a number of UPR-related genes and proteins are directly altered and related to the course of Prosurvival UPR in Cancer: The Example of ALS progression. GRP78/BiP, an ER chaperone, has been shown to be bound by SOD1 up-regulating its expression in Glioblastoma ALS in vivo models, and the calcium-binding chaperone GBM is the most frequent central nervous system primary calreticulin leads to motor neuron apoptosis.70 To further tumor. Its incidence is two to three new cases per 100,000

58 Doultsinos et al. 791 per year, and it has an extremely low prevalence; the The UPR as a Therapeutic Target: 5-year survival levels are less than 3%, resulting from a From the Tools to the Small Molecules dismal median prognosis of 15 months survival postdiag- nosis. GBM is a solid tumor–occupying lesion and there- The first report of an ER stress–induced adaptive response fore carries trademark characteristics such as poor in the yeast Saccharomyces cerevisiae was made in the 87 vascularization and high proliferation rates.78 The result- late 1980s using molecules perturbing protein folding in ing subsequent low pH and lack of nutrients offer major this compartment. As such, the alteration of protein triggers of potentially proapoptotic UPR involvement. N-glycosylation, for instance, by replacing glucose with However, tumors manage to adapt to such environments 2-deoxyglucose or using tunicamycin (an antibiotic that by modulating the UPR.79 Molecular chaperones have prevents the generation of glycans), leads to the accumula- 88 been heavily implicated in disease progression in GBM. tion of improperly folded proteins in the ER. Similarly, 2+ GRP78/BiP (high proliferation, prolonged cell survival) thapsigargin specifically inhibits the SERCA Ca ATPases, has been shown to be up-regulated in a variety of cancers thus preventing the refilling of ER calcium stores, leading 89 including GBM and shown to decrease in the presence of to functional imbalance of this compartment. Other chemotherapeutic agents, resulting in an increase in neo- compounds such as dithiothreitol, which is a reducing plastic cell apoptosis.80 Another chaperone, GRP94, dis- agent, or brefeldin A, which reversibly disables ER to 88,90 plays similar properties to GRP78 and is linked to ROS Golgi transport, induce the UPR. None of these com- UPR activation.81 The current treatment of choice in GBM pounds were usable in a clinical context because of their involves maximal tumor resection followed by the STUPP lack of specificity and high toxicity. Attempts to bypass protocol that involves treatment with the alkylating agent such effects are indeed under way, and it has been reported temozolomide alongside radiotherapy.78 This has variable that mipsagargin, a prodrug analog of thapsigargin, did success in patient cohorts as some high-grade gliomas are display acceptable tolerability and favorable pharmacoki- 91 resistant to the effects of this particular course of chemo- netic profiles in patients with solid tumors. Despite the therapy and adjuvant radiotherapy.79 Modulation of the wide-reaching application of ER stressors and the discov- UPR to sensitize resistant tumors to existing therapies ery of some currently used in clinics such as bortezomib 92 could be an attractive therapeutic target in GBM. IRE1 is in multiple myeloma, it is evident that further character- a key player involved in the apoptotic switch; however, it ization of the UPR and its potential modulation is needed produces either adaptive or death signals via its RNase for more effective therapies, justifying the approaches activity and was shown to play key roles in GBM develop- described below. ment, with particular emphasis on tumor growth and neoan- 82 giogenesis. It is currently not clear how IRE1 downstream Screening Strategies signaling events integrate to yield those specific outcomes. Using a dominant negative approach in U87 cells, IRE1 Taking into account the increasing number of reports was found to contribute to neoangiogenesis, and its RNase incriminating the UPR in the development of multiple dis- activity was involved in migration/invasion as well as in eases, targeting this signaling pathway is becoming a cen- proinflammatory processes.83 Silencing IRE1 RNase activ- tral public health and economic issue for numerous ity in glioblastoma cells pushed them toward a more motile laboratories and companies. The diversity of the in vitro and phenotype, showing short-range infiltration at the immedi- in vivo approaches employed to reach this goal can give a ate periphery of GBM cores and an extensive blood vessel picture of the inherent complexity of this signaling path- cooption with formation of distal perivascular tumor micro- way. In Table 1, we provide a nonexhaustive review of the satellites.82,83 It has been shown in orthotopic xenograft diversity of screening approaches that have led to the dis- models of human glioma that invalidation of both the kinase covery of new UPR modulators. In some cases (e.g., cancer and endoribonuclease domains of IRE1 produced avascular treatment), molecules that induce the UPR by initiating ER infiltrative tumors with vessel cooption, while invalidation stress could represent a way of overloading cellular resis- of the endoribonuclease function alone did not affect angio- tance to chemotherapy.120 In that case, screening strategies genesis, suggesting that the IRE1 domains play specific employing cell protection from ER stress–induced cell roles in the migration and vascularization of tumors.83 death or GRP78-luciferase reporter as readout have led to Moreover, it has been shown that IRE1 plays an important the identification of new specific ER stressors. Consequently, role in the up-regulation of extracellular matrix proteins drugs that inhibit a specific branch or protein of the UPR such as SPARC, further cementing IRE1’s involvement in could represent a relevant targeted therapy, first, in cancers tumor growth and migration.82,84,85 Finally, IRE1 has been whose development appears to be dependent of specific implicated in the maintenance of the circadian clock gene branches of the UPR121 or, second, in the case of treatment PER1, directly involving it in the survival and growth of against viruses that use the secretory pathway to produce tumor cells.86 difficult-to-fold viral proteins. Identification of such drugs

59 792 SLAS Discovery 22(7)

Table 1. Screening Approaches That Have Led to the Discovery of New Unfolded Protein Response (UPR) Modulators.

Target/ Pathway Screening Method Library Target Molecule(s) Reference UPR GRP78-luciferase reporter NA NA Versipelostatin 93 GRP78-luciferase reporter Cultured broth of NA Verrucosidin 94 microorganisms 19,000 compounds 95 GRP78-luciferase reporter 10,000 compounds NA BIX 96 NA 435 herbal medicine varieties NA Ponciri fructus 97 Comprehensive C. elegans feeding- C. elegans RNAi library obtained NA RNAi 98 mediated RNAi analysis from Source BioScience GRP78-luciferase reporter 20,000 commercial and NA Hydroxynaphthoic acids 99 in-house compounds IRE1 XBP1-luciferase reporter Cultured broth of NA Trierixin, mycotrienin II, 100 microorganisms and trienomycin A XBP1-luciferase reporter Cultured broth of NA Quinotrierixin, 101 microorganisms demethyltrienomycin A, and mycotrienin In vitro assay for IRE1 RNase NA IRE1 APY29, H6, imatinib, and 102 activity; fluorescence quenching sunitinib In vitro IRE1 RNase activity; Kinase inhibitor library IRE1 Flavonol quercetin 103 fluorescence quenching (BioMol International) In vitro oligomerization assay using NA IRE1 Peptide (FIRE) 104 AlphaScreen In vitro IRE1 RNase activity; 238,287 commercially available IRE1 4μ8c 105 fluorescence quenching chemicals and fractionated natural plant and fungal extracts In vitro IRE1 RNase activity; MannKind chemical library of IRE1 Salicylaldehyde analogs 106 fluorescence quenching 220,000 individual compounds XBP1-luciferase reporter Cultured broth of IRE1 Toyocamycin 107 microorganisms NA NA NA MKC-3946 108 XBP1-luciferase reporter 60,000 compounds (Chemdiv) IRE1 STF-083010 109 IRE1-PERK XBP1-luciferase and CHOP- 66,000 compounds + collection NA Patulin 110 luciferase reporters of 5036 natural extracts PERK In vitro Perk-mediated eIF2D Proprietary collection of PERK GSK2606414 111 phosphorylation; homogeneous kinase inhibitors time-resolved fluorescence ATF4-luciferase and XBP1- 106,281 compounds eIF2B ISRIB 112 luciferase reporters ATF6 ERSE-luciferase reporter and ATF6 106,281 compounds ATF6 Ceapins 113 nuclear translocation pathway ERAD Class I MHC–EGFP fusion protein 16,320 compounds of the p97/VCP Eeyarestatin 114 reporter Chembridge compound library GRP94 In vitro fluorescence polarization 130 purine-scaffold compounds GRP94 PU-H54, PU-WS13, and 115 assay PU-H39 ER folding Reporters of free chaperone content NA NA Azoramide 116 capacity and protein-folding capacity of the endoplasmic reticulum PDI Insulin-based turbidimetric assay 4900 compounds, including PDI Quercetin-3-rutinoside 117 modified for high-throughput approximately 3000 known screening bioactive compounds Glycoprotein Glycosylation-sensitive 2802 compounds from the NA Aclacinomycin 118 secretion luciferase reporters National Cancer Institute (NCI) chemical libraries CHOP CHOP-luciferase reporter 3000-compound space-filling Potentially TGD31BZ and 119 combinatorial library CYP51 TGD45BZ

60 Doultsinos et al. 793 has been dependent on the development of in vitro method- 384-well format and the monitoring of XBP1 activation ologies or cellular reporters of the IRE1, PERK, or ATF6 through an UPR pathway element-Luc–based assay in a pathways, using the specificity of their molecular activation 1536-well format.124 HTS also identified 2,9-diazaspiro[5.5] (XBP1 splicing, ATF4 translation, and ATF6 nuclear trans- undecanes as cytotoxicity inducers in several cell lines location, respectively; Table 1). In contrast, other diseases, through the depletion of intracellular calcium stores.125 The such as human amyloid diseases, are related to insufficient UPR inducer borrelidin was also discovered through a cell UPR. In such diseases, screening studies have mainly HTS in which UPR-inducing natural extracts were purified focused on the identification of UPR activators. In general, by RP-18 high-performance liquid chromatography and both in vitro and cellular strategies are employed in screen- iterative bioassay–guided C18 fractionation.113 ing strategies aiming to identify UPR modulators. Although cellular strategies are usually coupled with efforts to deter- Reporters of ER Functions mine the specificity of the hits, it is important to note that these approaches have enabled the identification of multi- Adding to the strategies employed to target the adaptation target drugs, which are particularly relevant in bypassing pathway of the UPR, we wanted to underline the efforts that neoplastic chemoresistance driven by mutation or compen- have been made to identify drugs that target the molecular satory mechanism processes. Finally, although in vivo mod- functions of the secretory pathway. In 2004, Fiebiger and col- els, such as Caenorhabditis elegans and Drosophila leagues114 established a fluorescent reporter of (ER)-to-cytosol melanogaster, provide promising platforms for efficient degradation pathway. They used a double construction consist- drug discovery and drug target identification, their use in ing of a class I major histocompatibility complex (MHC) high-throughput screening (HTS) studies of the UPR are up heavy chain–enhanced green fluorescent protein fusion protein to now underrepresented. (EGFP-HC) and the human cytomegalovirus protein US11. When both expressed in cells, US11 induces the dislocation of Cellular Reporters of UPR Activation MHC class I heavy chains from the ER to the cytosol, leading to its degradation, producing a weak fluorescence signal. Cell luminescent reporters have been widely used to iden- Using this system, the authors performed an HTS and identi- tify UPR modulators. These reporters have been designed fied two compounds, eeyarestatin I and II, which inhibit degra- to target (1) transcriptional activation of the UPR, by fusing dation of three dislocation substrates by retaining them in the luciferase genes to regulatory sequences such as GRP78/ ER. These were EGFP-HC, wild-type class I heavy chain and BiP or CHOP promoters or ER stress response element T-cell receptor. Bennet et al.118 adapted a glycosylation- sequences; (2) IRE1 activity, by cloning luciferase sequence sensitive luciferase reporter by the addition of the EGFR downstream of the 26-nt ER stress–specific intron of human N-terminal amino acid leader sequence to the luciferase gene XBP1; and (3) PERK activity, by fusing the 5′ UTR of the and the incorporation of glycosylation consensus sites into the human ATF4 containing the μORF to luciferase. Although luciferase coding sequence. Through the screen of 2802 com- these last methodologies were designed to screen for drugs pounds, the authors identified aclacinomycin as a compound that target one UPR branch, they usually require a second that reduces cell surface expression of glycosylated proteins. set of experiments to establish drug specificity. A study con- Fu and colleagues116 developed two HTS functional screening ducted by Walter and colleagues122 can be cited as a repre- systems that independently measure the free chaperone con- sentative experimental design of screening studies of the tent and protein-folding capacity of the ER. The first system UPR. By screening 106,281 compounds using an ATF4- consists of the fusion of a human gene fragment that encodes a luciferase reporter system, they initially identified 460 hits. peptide derived from the ATF6 luminal domain to luciferase. In a second phase, they discarded inhibitors that also This system allowed the measurement of the free chaperone affected the IRE1 branch of the UPR, by using a reporter of content in the ER. In conditions where ATF6 is released from XBP1 splicing, restricting the hits number to 187. This GRP78 (e.g., ER stress), the luciferase secretion is triggered, number was then reduced to 77 in a microscopy-based resulting in luminescence increase. The second system was screen that employed another ATF4 reporter. Subsequently, constructed by associating the coding sequence of the mem- Western blot analysis was used in a quaternary screen. brane protein asialoglycoprotein receptor 1 (ASGR1) to lucif- Finally, 28 compounds were selected for their ability to spe- erase gene. Considering that the ASGR1 expression is sensitive cifically target the PERK branch of the UPR and led to the to ER function, this reporter monitors ER-folding capacities. identification of ISRIB, a potent inhibitor of PERK signal- Using these cellular reporters, the authors identified a small- ing. A similar approach was used by the same group to iden- molecule compound, azoramide, that regulates ER folding and tify specific modulators of the ATF6 branch.123 Other secretion capacity without inducing ER stress and protects approaches to identify selective modulators of IRE1 have cells from the consequences of ER stress. This drug was pro- included the use of a chimeric CHOP-Gal4 transcription posed as a potential drug candidate for type 2 diabetes. These factor to monitor IRE1-mediated p38MAPK activation in a studies underline the multiplicity of processes that can be

61 794 SLAS Discovery 22(7) targeted in order to identify modulators of the ER functions priming them for activation.21,138,139 The understanding of the (reviewed in ref. 121). The CHOP pathway has been targeted physiology of these targets forms the basis of their potential to identify compounds that promote proapoptotic CHOP- pharmacological modulation. Modulators have been synthe- related cascades without triggering adaptive cascades. Through sized to both activate and inhibit IRE1. Activators of IRE1 HTS, an optimized sulfonamidebenzamide compound was predominantly act through the kinase domain by stabilizing a discovered that displayed antiproliferative effects spanning conformation that mimics the phosphorylated from of multiple cancer cell lines.126 IRE1.139 Compound 3 is one such molecule shown to inhibit IRE1 kinase activity while promoting at the same time XBP1 140 In Vitro Strategies splicing. APY-29 and sunitinib displayed similar functions, filling the adenine binding site of the kinase domain with high Fluorescence quenching–based, Alphascreen, or homoge- affinity, leading to partial molecular activation of IRE1.30 neous time-resolved fluorescence, have been proposed as tools Compounds have also been explored acting upstream of for the in vitro identification of modulators of central proteins IRE1, targeting BiP. A series of molecules (thiazole benzen- of the UPR such as IRE1, PERK, or GRP94 (Table 2). Note sulfonamides) have been shown to have antineoplastic prop- that these approaches are particularly relevant in the case of erties in melanoma cells by inducing autophagic and apoptotic some highly regulated proteins such as chaperones, taking into cell death.108 MKC-3946 (a noncompetitive inhibitor) of the account (1) that molecular processes could compensate for IRE1 RNAse site has been shown to inhibit growth of MM their inhibition and (2) the need of a high degree of specificity cell lines, avoiding toxicity in mononuclear cells.141 Moreover, due to the existence of highly homologous paralogs. These in bortezomib-induced cytotoxicity was shown to be increased vitro strategies led to the identification of highly selective in the presence of MKC, and ER stress induction by bortezo- modulators. Furthermore, they were also often followed by mib, deduced by the presence of XBP1s, was blocked by optimization studies aiming at improving drug affinity, physi- MKC-3946.142 Apoptosis induced by these agents was cochemical properties, metabolism, and so forth. As a result, enhanced by MKC-3946, associated with an increase in GSK2656157 was discovered as a decreased lipophilicity CHOP expression. Furthermore, MKC-3946 was shown to

GSK2606414 derivative, ISRIB underwent EC50 improve- inhibit XBP1 splicing in vivo, while inhibiting the growth of ment and optimization, IPA was engineered as a better IRE1 multiple myeloma cells.143 UPRM8 is a covalently bound activator than APY29, and BnIm was shown to present better RNAse inhibitor.144 The development of kinase inhibitors potency and selectivity for Grp94 than previously described involving the manipulation of an analogous DFG+2 nucleo- purine-scaffold compounds (Table 1). N-acridine-9-yl-N′,N′- phile (C715) located allosterically to the ATP binding site has dimethylpropane-1,3-diamine or DAPA and its derivatives been postulated to enable the identification of IRE1 inhibitors such as N9-(3-(dimethylamino)propyl)-N3,N3,N6,N6-tetra- that could overcome traditional specificity concerns. Chronic methylacridine-3,6,9-triamine (3,6-DMAD) were also identi- lymphocytic leukemia and non–small-cell lung cancer medi- fied as IRE1 inhibitors through an HTS chemical library screen cations ibrutinib and afatinib have set some precedence in the and were shown to be cytotoxic to multiple myeloma by inhib- clinical use of cysteine-directed covalent inhibitors, and iting both IRE1α oligomerization and XBP1 splicing.136 therefore, UPRM8 shows promise in its use as an IRE1 mod- Moreover, a dissociation-enhanced lanthanide fluorescence ulator in GBM.144,145 The kinase-inhibiting RNase attenuators immunoassay can be used to identify IRE1 autophosphoryla- KIRA3 and 6 act on the RNAse domain of IRE1 indirectly tion inhibitors by designing a 384-well HTS specific to this and were discovered during the screening of type II ATP com- branch of the UPR to produce drugs targeting multiple petitive ligands. KIRA3 (an optimized pyrazolopyrimidine myeloma and other secretory diseases.137 This approach was structure) specifically directly antagonizes the effect of IRE1 used to screen a library of 2312 potential kinase inhibitors, activation by APY29. KIRA3 favors the prevalence of the identifying 30 compounds that were shown to bind to the IRE1 monomeric inactive form of IRE1, and although effective, kinase domain, including some known inhibitors such as suni- KIRA6 (a KIRA3 derivative; Table 2) was developed as a tinib and some unknown such as CCT249525.137 more thermodynamically stable alternative for use in in vivo models. KIRAs offer an alternative to modulation with sali- Molecules Targeting IRE1 and Their cylaldehyde inhibitors such as MKC as they inhibit not only Use in Disease Models the RNAse but also kinase domains of IRE1, offering a more complete blocking alternative.146 Recent attempts to obtain The detection of the accumulated unfolded proteins is an novel tricyclic chromenone-based inhibitors have been pre- essential initial step for the activation of the three arms of the sented as useful tools to study XBP1 suppression in whole UPR. This detection is mediated by GRP78 or BiP, which, cells. Inhibitors of the peptidyl-propyl isomerases and HSP90 under normal physiological conditions, is bound to the lumi- inhibitors such as geldanamycin have been developed pri- nal domain of the sensors, thereby keeping them deactivated. marily targeting immunosuppression and anticancer treat- ER stress causes the dissociation of BiP from the sensors, ment. This has followed the association of such inhibitors

62 Doultsinos et al. 795

Table 2. Molecules Identified to Target Directly the ATF6, IRE1, or PERK Pathways.

Pathway Name Structure Target Reference ATF6 16F16 PDI 47

Ceapins ATF6D 113

Compound 147 ATF6D 127

Compound 263 ATF6D 127

IRE1 4μ8c IRE1 RNase 105

MKC analogues IRE1 RNase 106, 128–130

3-ethoxy-5,6-dibromosalicylaldehyde IRE1 RNase 106

KIRA6 IRE1 kinase 131

STF-083010 IRE1 RNase 109

Compound 3 IRE1 kinase 132

PERK GSK2656157 PERK 133

ISRIB eIF2E 122

Salubrinal GADD34 (PP1c) 134

Guanabenz GADD34 (PP1c) 134

Sephin 1/IFB-088 GADD34 (PP1c) 135

63 796 SLAS Discovery 22(7) with the HSP90 ER paralog GRP94 and evidence pointing Funding toward the HSP90 dependent destabilization of UPR signal- The authors disclosed receipt of the following financial support 147 ing. In a different study, 11 overlapping fragments of the for the research, authorship, and/or publication of this article: This IRE1 cytosolic domain (F1-F11) were generated as 6xHis work was funded by grants from the Institut National du Cancer fusion proteins to test their effect on IRE1 signaling. One of (INCa) and EU H2020 MSCA ITN-675448 (TRAINERS). those, F6, its peptide derivative P4, and its modified version TAT-P4 (FIRE), have been shown to promote ER stress resis- References tance by reducing RIDD and JNK activation and enhancing 1. Palade, G. E. The Endoplasmic Reticulum. J. Biophys. XBP1s production. However, the exploration of peptidomi- Biochem. Cytol. 1956, 2, 85–98. metic design targeting pharmacological chaperones and pro- 2. Voeltz, G. K.; Rolls, M. M.; Rapoport, T. A. Structural tein folding control remains to be fully addressed.104 Organization of the Endoplasmic Reticulum. EMBO Rep. Modulators have also targeted the other two arms of the 2002, 3, 944–950. UPR. ISRIB has been shown to be a potent inhibitor of the 3. Zhang, K.; Wang, S.; Malhotra, J.; et al. The Unfolded Protein integrated stress response by modulating eIF2α phosphoryla- Response Transducer IRE1alpha Prevents ER Stress-Induced tion and its translational effects.112 Salubrinal and guanabenz Hepatic Steatosis. EMBO J. 2011, 30, 1357–1375. inhibit the GADD34/PP1c complex, thereby preventing eIF2α 4. Michalak, M.; Robert Parker, J. M.; Opas, M. Ca2+ Signaling and Calcium Binding Chaperones of the Endoplasmic dephosphorylation, although the direct or indirect mechanism Reticulum. Cell Calcium 2002, 32, 269–278. by which this is achieved is not yet clear and is currently the 5. Kamhi-Nesher, S.; Shenkman, M.; Tolchinsky, S.; et al. subject of ongoing investigation. The GADD34/PP1c axis is A Novel Quality Control Compartment Derived from the also targeted by sephin1/IFB-088. In addition, GSK2656157 Endoplasmic Reticulum. Mol. Biol. Cell 2001, 12, 1711– 95,128,134 inhibits PERK directly. Modulators targeting the ATF6 1723. pathway include the newly identified ceapins that trap ATF6α 6. Zuber, C.; Spiro, M. J.; Guhl, B.; et al. Golgi Apparatus in ER-resident foci, thereby inhibiting its translocation to the Immunolocalization of Endomannosidase Suggests Post- Golgi and compound 147 that preferentially targets the ATF6 Endoplasmic Reticulum Glucose Trimming: Implications for transcriptional repertoire (Table 2).123,127 Quality Control. Mol. Biol. Cell 2000, 11, 4227–4240. 7. Zhang, K.; Kaufman, R. J. Signaling the Unfolded Protein Response from the Endoplasmic Reticulum. J. Biol. Chem. Conclusions 2004, 279, 25935–25938. 8. Ellgaard, L.; Helenius, A. Quality Control in the Endoplasmic In this review, we have illustrated the critical functions Reticulum. Nat. Rev. Mol. Cell Biol. 2003, 4, 181–191. played by the regulation of ER homeostasis in health and 9. Jin, H.; Yan, Z.; Nam, K. H.; et al. Allele-Specific diseases, with particular attention to three examples of Suppression of a Defective Brassinosteroid Receptor Reveals pathologies comprising ALS, diabetes, and cancer (glio- a Physiological Role of UGGT in ER Quality Control. Mol. blastoma). We also discuss the relevance of targeting the Cell 2007, 26, 821–830. 10. Taylor, S. C.; Ferguson, A. D.; Bergeron, J. J.; et al. The main ER homeostasis control pathway, namely, the UPR, in ER Protein Folding Sensor UDP-Glucose Glycoprotein- diseases either to improve an altered ER homeostasis and Glucosyltransferase Modifies Substrates Distant to Local rescue cells from death (e.g., for ALS) or to weaken cells’ Changes in Glycoprotein Conformation. Nat. Struct. Mol. adaptive capacity to enhance death (e.g., cancer/GBM). Biol. 2004, 11, 128–134. Screening strategies have identified selective modulators of 11. Schrag, J. D.; Bergeron, J. J.; Li, Y.; et al. The Structure of the UPR sensors for which a clinical future can be envi- Calnexin, an ER Chaperone Involved in Quality Control of sioned. On the other hand, clinically relevant compounds Protein Folding. Mol. Cell 2001, 8, 633–644. with applications in various cancers, such as the renal, 12. Nishikawa, S.; Brodsky, J. L.; Nakatsukasa, K. Roles of hepatocellular, and thyroid carcinoma (sorafenib) or mela- Molecular Chaperones in Endoplasmic Reticulum (ER) noma (vemurafenib), have been shown to mediate their Quality Control and ER-Associated Degradation (ERAD). J. toxic effects in part through the perturbation of ER homeo- Biochem. 2005, 137, 551–555. 13. Schroder, M.; Kaufman, R. J. The Mammalian Unfolded stasis and activation of the UPR. As a consequence, the Protein Response. Annu. Rev. Biochem. 2005, 74, 739–789. value of the UPR as a therapeutic target in diseases with 14. Hetz, C.; Chevet, E.; Oakes, S. A. Proteostasis Control by the dismal prognoses continues to be proven. Further investiga- Unfolded Protein Response. Nat. Cell Biol. 2015, 17, 829–38. tions can only show how wide ranging the reach of UPR 15. Hitomi, J.; Katayama, T.; Eguchi, Y.; et al. Involvement targeting therapeutics can be. of Caspase-4 in Endoplasmic Reticulum Stress-Induced Apoptosis and Abeta-Induced Cell Death. J. Cell Biol. 2004, 165, 347–356. Declaration of Conflicting Interests 16. Tirasophon, W.; Welihinda, A. A.; Kaufman, R. J. A Stress The authors declared no potential conflicts of interest with respect Response Pathway from the Endoplasmic Reticulum to the to the research, authorship, and/or publication of this article. Nucleus Requires a Novel Bifunctional Protein Kinase/

64 Doultsinos et al. 797

endoribonuclease (Ire1p) in Mammalian Cells. Genes Dev. 34. Maurel, M.; Chevet, E.; Tavernier, J.; et al. Getting RIDD 1998, 12, 1812–1824. of RNA: IRE1 in Cell Fate Regulation. Trends Biochem. Sci. 17. Adachi, Y.; Yamamoto, K.; Okada, T.; et al. ATF6 Is a 2014, 39, 245–254. Transcription Factor Specializing in the Regulation of Quality 35. Harding, H. P.; Zhang, Y.; Bertolotti, A.; et al. Perk Is Control Proteins in the Endoplasmic Reticulum. Cell Struct. Essential for Translational Regulation and Cell Survival Funct. 2008, 33, 75–89. during the Unfolded Protein Response. Mol. Cell 2000, 5, 18. B’Chir, W.; Maurin, A. C.; Carraro, V.; et al. The eIF2α/ 897–904. ATF4 Pathway Is Essential for Stress-Induced Autophagy 36. Wek, R. C.; Jiang, H.-Y.; Anthony, T. G. Coping with Stress: Gene Expression. Nucleic Acids Res. 2013, 41, 7683–7699. eIF2 Kinases and Translational Control. Biochem. Soc. Trans. 19. Bertolotti, A.; Zhang, Y.; Hendershot, L. M.; et al. Dynamic 2006, 34, 7–11. Interaction of BiP and ER Stress Transducers in the Unfolded- 37. Ameri, K.; Harris, A. L. Activating Transcription Factor 4. Protein Response. Nat. Cell Biol. 2000, 2, 326–332. Int. J. Biochem. Cell Biol. 2008, 40, 14–21. 20. Kimata, Y.; Oikawa, D.; Shimizu, Y.; et al. A Role for BiP 38. Harding, H. P.; Zhang, Y.; Zeng, H.; et al. An Integrated as an Adjustor for the Endoplasmic Reticulum Stress-Sensing Stress Response Regulates Amino Acid Metabolism and Protein Ire1. J. Cell Biol. 2004, 167, 445–456. Resistance to Oxidative Stress. Mol. Cell 2003, 11, 619–633. 21. Shen, J.; Chen, X.; Hendershot, L.; et al. ER Stress Regulation 39. Roybal, C. N.; Yang, S.; Sun, C. W.; et al. Homocysteine of ATF6 Localization by Dissociation of BiP/GRP78 Binding Increases the Expression of Vascular Endothelial Growth and Unmasking of Golgi Localization Signals. Dev. Cell Factor by a Mechanism Involving Endoplasmic Reticulum 2002, 3, 99–111. Stress and Transcription Factor ATF4. J. Biol. Chem. 2004, 22. Oikawa, D.; Kimata, Y.; Kohno, K.; et al. Activation 279, 14844–14852. of Mammalian IRE1alpha upon ER Stress Depends on 40. Todd, D. J.; Lee, A. H.; Glimcher, L. H. The Endoplasmic Dissociation of BiP Rather Than on Direct Interaction with Reticulum Stress Response in Immunity and Autoimmunity. Unfolded Proteins. Exp. Cell Res. 2009, 315, 2496–2504. Nat. Rev. Immunol. 2008, 8, 663–674. 23. Pincus, D.; Chevalier, M. W.; Aragon, T.; et al. BiP Binding to 41. Novoa, I.; Zeng, H.; Harding, H. P.; et al. Feedback Inhibition the ER-Stress Sensor Ire1 Tunes the Homeostatic Behavior of of the Unfolded Protein Response by GADD34-Mediated the Unfolded Protein Response. PLoS Biol. 2010, 8, e1000415. Dephosphorylation of eIF2alpha. J. Cell. Biol. 2001, 153, 24. Hetz, C.; Chevet, E.; Harding, H. P. Targeting the Unfolded 1011–1022. Protein Response in Disease. Nat. Rev. Drug Discov. 2013, 42. Eletto, D.; Eletto, D.; Boyle, S.; et al. PDIA6 Regulates 12, 703–719. Insulin Secretion by Selectively Inhibiting the RIDD Activity 25. Tirasophon, W.; Lee, K.; Callaghan, B.; et al. The of IRE1. FASEB J. 2016, 30, 653–665. Endoribonuclease Activity of Mammalian IRE1 Autoregulates 43. Nioi, P.; McMahon, M.; Itoh, K.; et al. Identification of a Its mRNA and Is Required for the Unfolded Protein Response. Novel Nrf2-Regulated Antioxidant Response Element (ARE) Genes Dev. 2000, 14, 2725–2736. in the Mouse NAD(P)H:quinone Oxidoreductase 1 Gene: 26. Martino, M. B.; Jones, L.; Brighton, B.; et al. The ER Stress Reassessment of the ARE Consensus Sequence. Biochem. J. Transducer IRE1β Is Required for Airway Epithelial Mucin 2003, 374, 337–348. Production. Mucosal Immunol. 2013, 6, 639–654. 44. Chen, X.; Shen, J.; Prywes, R. The Luminal Domain of ATF6 27. Tsuru, A.; Fujimoto, N.; Takahashi, S.; et al. Negative Senses Endoplasmic Reticulum (ER) Stress and Causes Feedback by IRE1β Optimizes Mucin Production in Goblet Translocation of ATF6 from the ER to the Golgi. J. Biol. Cells. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, 2864–2869. Chem. 2002, 277, 13045–13052. 28. Hetz, C. Protein Misfolding Disorders: A Trip into the ER. 45. Lee, K.; Tirasophon, W.; Shen, X.; et al. IRE1-Mediated Potomac, MD: Bentham eBooks; 2009. Unconventional mRNA Splicing and S2P-Mediated ATF6 29. Liu, C. Y.; Wong, H. N.; Schauerte, J. A.; et al. The Protein Cleavage Merge to Regulate XBP1 in Signaling the Unfolded Kinase/Endoribonuclease IRE1alpha That Signals the Protein Response. Genes Dev. 2002, 16, 452–466. Unfolded Protein Response Has a Luminal N-Terminal 46. Shoulders, M. D.; Ryno, L. M.; Genereux, J. C.; et al. Stress- Ligand-Independent Dimerization Domain. J. Biol. Chem. Independent Activation of XBP1s And/or ATF6 Reveals 2002, 277, 18346–18356. Three Functionally Diverse ER Proteostasis Environments. 30. Korennykh, A. V; Egea, P. F.; Korostelev, A. A.; et al. The Cell Rep. 2013, 3, 1279–1292. Unfolded Protein Response Signals through High-Order 47. Higa, A.; Taouji, S.; Lhomond, S.; et al. Endoplasmic Assembly of Ire1. Nature 2009, 457, 687–693. Reticulum Stress-Activated Transcription Factor ATF6alpha 31. Lu, Y.; Liang, F. X.; Wang, X. A Synthetic Biology Approach Requires the Disulfide Isomerase PDIA5 to Modulate Identifies the Mammalian UPR RNA Ligase RtcB. Mol. Cell Chemoresistance. Mol. Cell Biol. 2014, 34, 1839–1849. 2014, 55, 758–770. 48. Li, G.; Mongillo, M.; Chin, K. T.; et al. Role of ERO1- 32. Hollien, J.; Lin, J. H.; Li, H.; et al. Regulated Ire1-Dependent Alpha-Mediated Stimulation of Inositol 1,4,5-Triphosphate Decay of Messenger RNAs in Mammalian Cells. J. Cell. Biol Receptor Activity in Endoplasmic Reticulum Stress-Induced 2009, 186, 323–331. Apoptosis. J. Cell Biol. 2009, 186, 783–792. 33. Eletto, D.; Dersh, D.; Gidalevitz, T.; et al. Protein Disulfide 49. Urra, H.; Dufey, E.; Lisbona, F.; et al. When ER Stress Isomerase A6 Controls the Decay of IRE1alpha Signaling via Reaches a Dead End. Biochim. Biophys. Acta 2013, 1833, Disulfide-Dependent Association. Mol. Cell 2014, 53, 562–576. 3507–3517.

65 798 SLAS Discovery 22(7)

50. de Mattia, F.; Gubser, C.; van Dommelen, M. M.; et al. 67. Hetz, C. The Unfolded Protein Response: Controlling Cell Human Golgi Antiapoptotic Protein Modulates Intracellular Fate Decisions under ER Stress and beyond. Nat. Rev. Mol. Calcium Fluxes. Mol. Biol. Cell 2009, 20, 3638–3645. Cell Biol. 2012, 13, 89–102. 51. Hetz, C.; Russelakis-Carneiro, M.; Maundrell, K.; et al. 68. Turner, B. J.; Talbot, K. Transgenics, Toxicity and Caspase-12 and Endoplasmic Reticulum Stress Mediate Therapeutics in Rodent Models of Mutant SOD1-Mediated Neurotoxicity of Pathological Prion Protein. EMBO J. 2003, Familial ALS. Prog. Neurobiol. 2008, 85, 94–134. 22, 5435–5445. 69. Barmada, S. J.; Serio, A.; Arjun, A.; et al. Autophagy Induction 52. Matsuzaki, S.; Hiratsuka, T.; Kuwahara, R.; et al. Caspase-4 Enhances TDP43 Turnover and Survival in Neuronal ALS Is Partially Cleaved by Calpain via the Impairment of Ca2+ Models. Nat. Chem. Biol. 2014, 10, 677–685. Homeostasis under the ER Stress. Neurochem. Int. 2009, 56, 70. Bernard-Marissal, N.; Moumen, A.; Sunyach, C.; et al. 352–356. Reduced Calreticulin Levels Link Endoplasmic Reticulum 53. Asada, R.; Kanemoto, S.; Kondo, S.; et al. The Signalling Stress and Fas-Triggered Cell Death in Motoneurons from Endoplasmic Reticulum-Resident bZIP Transcription Vulnerable to ALS. J. Neurosci. 2012, 32, 4901–4912. Factors Involved in Diverse Cellular Physiology. J. Biochem. 71. Hetz, C.; Mollereau, B. Disturbance of Endoplasmic 2011, 149, 507–518. Reticulum Proteostasis in Neurodegenerative Diseases. Nat. 54. Kondo, S.; Saito, A.; Asada, R.; et al. Physiological Unfolded Rev. Neurosci. 2014, 15, 233–249. Protein Response Regulated by OASIS Family Members, 72. Jaronen, M.; Vehviläinen, P.; Malm, T.; et al. Protein Disulfide Transmembrane bZIP Transcription Factors. IUBMB Life Isomerase in ALS Mouse Glia Links Protein Misfolding with 2011, 63, 233–239. NADPH Oxidase-Catalyzed Superoxide Production. Hum. 55. Kondo, S.; Saito, A.; Hino, S.; et al. BBF2H7, a Novel Mol. Genet. 2013, 22, 646–655. Transmembrane bZIP Transcription Factor, Is a New Type 73. World Health Organization. Global Report on Diabetes. 2016. of Endoplasmic Reticulum Stress Transducer. Mol. Cell Biol. http://apps.who.int/iris/bitstream/10665/204871/1/978924 2007, 27, 1716–1729. 1565257_eng.pdf. 56. Vecchi, C.; Montosi, G.; Zhang, K.; et al. ER Stress Controls 74. Qiu, Y.; Mao, T.; Zhang, Y.; et al. A Crucial Role for RACK1 in Iron Metabolism through Induction of Hepcidin. Science the Regulation of Glucose-Stimulated IRE1alpha Activation in 2009, 325, 877–880. Pancreatic Beta Cells. Sci. Signal 2010, 3, 106, ra7. 57. Ye, Y.; Meyer, H. H.; Rapoport, T. A. The AAA ATPase 75. Lee, A.-H.; Heidtman, K.; Hotamisligil, G. S.; et al. Dual and Cdc48/p97 and Its Partners Transport Proteins from the ER Opposing Roles of the Unfolded Protein Response Regulated by into the Cytosol. Nature 2001, 414, 652–656. IRE1α and XBP1 in Proinsulin Processing and Insulin Secretion. 58. Caruso, M. E.; Jenna, S.; Bouchecareilh, M.; et al. GTPase- Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 8885–8890. Mediated Regulation of the Unfolded Protein Response in 76. Fonseca, S. G.; Gromada, J.; Urano, F. Endoplasmic Caenorhabditis elegans Is Dependent on the AAA+ ATPase Reticulum Stress and Pancreatic Beta Cell Death. Trends CDC-48. Mol. Cell Biol. 2008, 28, 4261–4274. Endocrinol. Metab. 2011, 22, 266–274. 59. Marza, E.; Taouji, S.; Barroso, K.; et al. Genome-Wide 77. Karali, E.; Bellou, S.; Stellas, D.; et al. VEGF Signals through Screen Identifies a Novel p97/CDC-48-Dependent Pathway ATF6 and PERK to Promote Endothelial Cell Survival and Regulating ER-Stress-Induced Gene Transcription. EMBO Angiogenesis in the Absence of ER Stress. Mol. Cell 2014, Rep. 2015, 16, 332–340. 54, 559–572. 60. Luo, B.; Lee, A. S. The Critical Roles of Endoplasmic 78. Stupp, R.; Mason, W. P.; van den Bent, M. J.; et al. Reticulum Chaperones and Unfolded Protein Response in Radiotherapy plus Concomitant and Adjuvant Temozolomide Tumorigenesis and Anticancer Therapies. Oncogene 2013, for Glioblastoma. N. Engl. J. Med. 2005, 352, 987–996. 32, 805–818. 79. Pyrko, P.; Schonthal, A. H.; Hofman, F. M.; et al. The 61. Kanekura, K.; Suzuki, H.; Aiso, S.; et al. ER Stress and Unfolded Protein Response Regulator GRP78/BiP as a Unfolded Protein Response in Amyotrophic Lateral Sclerosis. Novel Target for Increasing Chemosensitivity in Malignant Mol. Neurobiol. 2009, 39, 81–89. Gliomas. Cancer Res. 2007, 67, 9809–9816. 62. Dicks, N.; Gutierrez, K.; Michalak, M.; et al. Endoplasmic 80. Wang, M.; Wey, S.; Zhang, Y.; et al. Role of the Unfolded Reticulum Stress, Genome Damage, and Cancer. Front. Protein Response Regulator GRP78/BiP in Development, Oncol. 2015, 5, 1–8. Cancer, and Neurological Disorders. Antioxid. Redox Signal. 63. Ju, J. S.; Fuentealba, R. A.; Miller, S. E.; et al. Valosin- 2009, 11, 2307–2316. Containing Protein (VCP) Is Required for Autophagy and Is 81. Lee, A. S. Glucose-Regulated Proteins in Cancer: Molecular Disrupted in VCP Disease. J. Cell Biol. 2009, 187, 875–888. Mechanisms and Therapeutic Potential. Nat. Rev. Cancer 64. Mackenzie, I. R.; Rademakers, R.; Neumann, M. TDP-43 and 2014, 14, 263–76. FUS in Amyotrophic Lateral Sclerosis and Frontotemporal 82. Auf, G.; Jabouille, A.; Guerit, S.; et al. Inositol-Requiring Dementia. Lancet Neurol. 2010, 9, 995–1007. Enzyme 1alpha Is a Key Regulator of Angiogenesis and 65. Jiang, N.; Shen, Y.; Fei, X.; et al. Valosin-Containing Protein Invasion in Malignant Glioma. Proc. Natl. Acad. Sci. U.S.A. Regulates the Proteasome-Mediated Degradation of DNA- 2010, 107, 15553–15558. PKcs in Glioma Cells. Cell Death Dis. 2013, 4, e647. 83. Jabouille, A.; Delugin, M.; Pineau, R.; et al. Glioblastoma 66. Orrell, R. W. Understanding the Causes of Amyotrophic Invasion and Cooption Depend on IRE1α Endoribonuclease Lateral Sclerosis. N. Engl. J. Med. 2007, 357, 822–823. Activity. Oncotarget 2015, 6, 24922–24934.

66 Doultsinos et al. 799

84. Dejeans, N.; Pluquet, O.; Lhomond, S.; et al. Autocrine 100. Futamura, Y.; Tashiro, E.; Hironiwa, N.; et al. Trierixin, Control of Glioma Cells Adhesion and Migration through a Novel Inhibitor of ER Stress-Induced XBP1 Activation IRE1alpha-Mediated Cleavage of SPARC mRNA. J. Cell Sci. from Streptomyces Sp. II. Structure Elucidation. J. Antibiot. 2012, 125, 4278–4287. 2007, 60, 582–585. 85. Ma, Y.; Hendershot, L. M. The Role of the Unfolded Protein 101. Kawamura, T.; Tashiro, E.; Yamamoto, K.; et al. SAR Study of Response in Tumour Development: Friend or Foe? Nat. Rev. a Novel Triene-Ansamycin Group Compound, Quinotrierixin, Cancer 2004, 4, 966–977. and Related Compounds, as Inhibitors of ER Stress-Induced 86. Pluquet, O.; Dejeans, N.; Bouchecareilh, M.; et al. Posttranscrip- XBP1 Activation. J. Antibiot. 2008, 61, 303–311. tional Regulation of PER1 Underlies the Oncogenic Function of 102. Han, D.; Lerner, A. G.; Vande Walle, L.; et al. IRE1alpha Kinase IREalpha. Cancer Res. 2013, 73, 4732–4743. Activation Modes Control Alternate Endoribonuclease 87. Kozutsumi, Y.; Segal, M.; Normington, K.; et al. The Presence Outputs to Determine Divergent Cell Fates. Cell 2009, 138, of Malfolded Proteins in the Endoplasmic Reticulum Signals 562–575. the Induction of Glucose-Regulated Proteins. Nature 1988, 103. Wiseman, R. L.; Zhang, Y.; Lee, K. P.; et al. Flavonol Activation 332, 462–464. Defines an Unanticipated Ligand-Binding Site in the Kinase- 88. Oslowski, C. M.; Urano, F. Measuring ER Stress and the RNase Domain of IRE1. Mol. Cell 2010, 38, 291–304. Unfolded Protein Response Using Mammalian Tissue Culture 104. Bouchecareilh, M.; Higa, A.; Fribourg, S.; et al. Peptides System. Methods Enzymol. 2011, 490, 71–92. Derived from the Bifunctional kinase/RNase Enzyme 89. Thastrup, O. L. E.; Cullent, P. J.; Michael, R.; et al. IRE1{alpha} Modulate IRE1{alpha} Activity and Protect Thapsigargin, a Tumor Promoter, Discharges Intracellular Cells from Endoplasmic Reticulum Stress. FASEB J 2011, Ca2 + Stores by Specific Inhibition of the Endoplasmic 25, 3115–3129. Reticulum. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 2466– 105. Cross, B. C.; Bond, P. J.; Sadowski, P. G.; et al. The 2470. Molecular Basis for Selective Inhibition of Unconventional 90. Sciaky, N.; Presley, J.; Smith, C.; et al. Golgi Tubule Traffic mRNA Splicing by an IRE1-Binding Small Molecule. Proc. and the Effects of Brefeldin A Visualized in Living Cells. J. Natl. Acad. Sci. U.S.A. 2012, 109, E869–E878. Cell Biol. 1997, 139, 1137 LP–1155. 106. Volkmann, K.; Lucas, J. L.; Vuga, D.; et al. Potent and 91. Mahalingam, D.; Wilding, G.; Denmeade, S.; et al. Selective Inhibitors of the Inositol-Requiring Enzyme 1 Mipsagargin, a Novel Thapsigargin-Based PSMA-Activated Endoribonuclease. J. Biol. Chem. 2011, 286, 12743–12755. Prodrug: Results of a First-in-Man Phase I Clinical Trial 107. Ri, M.; Tashiro, E.; Oikawa, D.; et al. Identification of in Patients with Refractory, Advanced or Metastatic Solid Toyocamycin, an Agent Cytotoxic for Multiple Myeloma Tumours. Br. J. Cancer 2016, 114, 986–994. Cells, as a Potent Inhibitor of ER Stress-Induced XBP1 92. Weatherbee, J. L.; Kraus, J.; Ross, A. H. ER Stress in mRNA Splicing. Blood Cancer J. 2012, 2, e79. Temozolomide-Treated Glioblastomas Interferes with DNA 108. Joshi, A.; Newbatt, Y.; McAndrew, P. C.; et al. Molecular Repair and Induces Apoptosis. Oncotarget 2016, 7, 43820– Mechanisms of Human IRE1 Activation through Dimerization 43834. and Ligand Binding. Oncotarget 2015, 6, 13019–13035. 93. Park, H.-R.; Furihata, K.; Hayakawa, Y.; et al. Versipelostatin, 109. Papandreou, I.; Denko, N. C.; Olson, M.; et al. Identification a Novel GRP78/Bip Molecular Chaperone down-Regulator of of an Ire1alpha Endonuclease Specific Inhibitor with Microbial Origin. Tetrahedron Lett. 2002, 43, 6941–6945. Cytotoxic Activity against Human Multiple Myeloma. 94. Park, H.-R.; Ryoo, I.-J.; Choo, S.-J.; et al. Glucose-Deprived Blood 2011, 117, 1311–1314. HT-29 Human Colon Carcinoma Cells Are Sensitive to 110. Fribley, A. M.; Cruz, P. G.; Miller, J. R.; et al. Complementary Verrucosidin as a GRP78 down-Regulator. Toxicology 2007, Cell-Based High Throughput Screens Identify Novel 229, 253–261. Modulators of the Unfolded Protein Response. J. Biomol. 95. Boyce, M.; Bryant, K. F.; Jousse, C.; et al. A Selective Screen. 2011, 16, 825–835. Inhibitor of eIF2alpha Dephosphorylation Protects Cells from 111. Axten, J. M.; Medina, J. R.; Feng, Y.; et al. Discovery of ER Stress. Science 2005, 307, 935–939. 7-Methyl-5-(1-{[3-(Trifluoromethyl)phenyl]acetyl}-2,3-Di- 96. Kudo, T.; Kanemoto, S.; Hara, H.; et al. A Molecular hydro-1H-Indol-5-Yl)-7H-P yrrolo[2,3-D]pyrimidin-4-Amine Chaperone Inducer Protects Neurons from ER Stress. Cell (GSK2606414), a Potent and Selective First-in-Class Inhibitor Death Differ. 2008, 15, 364–375. of Protein Kinase R (PKR)-like Endoplasmic Reticulum Kinase 97. Cha, M.-R.; Yoon, M.; Son, E.-S.; et al. Selective Cytotoxicity (PERK). J. Med. Chem. 2012, 55, 7193–7207. of Ponciri Fructus against Glucose-Deprived PANC-1 Human 112. Sidrauski, C.; McGeachy, A. M.; Ingolia, N. T.; et al. The Pancreatic Cancer Cells via Blocking Activation of GRP78. Small Molecule ISRIB Reverses the Effects of eIF2al- Biosci. Biotechnol. Biochem. 2009, 73, 2167–2171. pha Phosphorylation on Translation and Stress Granule 98. Sakaki, K.; Yoshina, S.; Shen, X.; et al. RNA Surveillance Assembly. Elife 2015, 4, e05033. Is Required for Endoplasmic Reticulum Homeostasis. Proc. 113. Sidhu, A.; Miller, J. R.; Tripathi, A.; et al. Borrelidin Induces Natl. Acad. Sci. U.S.A. 2012, 109, 8079–8084. the Unfolded Protein Response in Oral Cancer Cells and 99. Jeong, K.; Ku, J.; Park, M.; et al. Hydroxynaphthoic Chop-Dependent Apoptosis. ACS Med. Chem. Lett. 2015, 6, Acids Identified in a High Throughput Screening Potently 1122–1127. Ameliorate Endoplasmic Reticulum Stress as Novel Chemical 114. Fiebiger, E.; Hirsch, C.; Vyas, J. M.; et al. Dissection of the Chaperones. Chem. Pharm. Bull. 2013, 61, 740–746. Dislocation Pathway for Type I Membrane Proteins with a

67 800 SLAS Discovery 22(7)

New Small Molecule Inhibitor, Eeyarestatin. Mol. Biol. Cell 131. Ghosh, R.; Likun, W.; Wang, E. S.; et al. Allosteric 2004, 15, 1635–1646. Inhibition of the IRE1α RNase Preserves Cell Viability and 115. Patel, P. D.; Yan, P.; Seidler, P. M.; et al. Paralog-Selective Function during Endoplasmic Reticulum Stress. Cell 2014, Hsp90 Inhibitors Define Tumor-Specific Regulation of 158, 534–548. Her2. Nat. Chem. Biol. 2013, 9, 677–684. 132. Wang, L.; Perera, B. G.; Hari, S. B.; et al. Divergent 116. Fu, S.; Yalcin, A.; Lee, G. Y.; et al. Phenotypic Assays Allosteric Control of the IRE1alpha Endoribonuclease Identify Azoramide as a Small-Molecule Modulator of the Using Kinase Inhibitors. Nat. Chem. Biol. 2012, 8, 982–989. Unfolded Protein Response with Antidiabetic Activity. Sci. 133. Atkins, C.; Liu, Q.; Minthorn, E.; et al. Characterization Transl. Med. 2015, 7, 292ra98 LP–292ra98. of a Novel PERK Kinase Inhibitor with Antitumor and 117. Jasuja, R.; Passam, F. H.; Kennedy, D. R.; et al. Protein Antiangiogenic Activity. Cancer Res. 2013, 73, 1993 LP-2002. Disulfide Isomerase Inhibitors Constitute a New Class of 134. Tsaytler, P.; Harding, H. P.; Ron, D.; et al. Selective Antithrombotic Agents. J. Clin. Invest. 2012, 122, 2104–2113. Inhibition of a Regulatory Subunit of Protein Phosphatase 1 118. Bennett, D. C.; Charest, J.; Sebolt, K.; et al. High-Throughput Restores Proteostasis. Science 2011, 332, 91–94. Screening Identifies Aclacinomycin as a Radiosensitizer of 135. Das, I.; Krzyzosiak, A.; Schneider, K.; et al. Preventing EGFR-Mutant Non-Small Cell Lung Cancer. Transl. Oncol. Proteostasis Diseases by Selective Inhibition of a Phosphatase 2013, 6, 382–391. Regulatory Subunit. Science 2015, 348, 239–242. 119. Harding, H. P.; Zhang, Y.; Khersonsky, S.; et al. Bioactive 136. Jiang, D.; Tam, A. B.; Alagappan, M.; et al. Acridine Small Molecules Reveal Antagonism between the Integrated Derivatives as Inhibitors of the IRE1α–XBP1 Pathway Are Stress Response and Sterol-Regulated Gene Expression. Cytotoxic to Human Multiple Myeloma. Mol. Cancer Ther. Cell Metab. 2005, 2, 361–371. 2016, 15, 2055 LP–2065. 120. Dejeans, N.; Manie, S.; Hetz, C.; et al. Addicted to Secrete— 137. Newbatt, Y.; Hardcastle, A.; McAndrew, P. C.; et al. Novel Concepts and Targets in Cancer Therapy. Trends Identification of Autophosphorylation Inhibitors of the Mol. Med. 2014, 20, 242–250. Inositol-Requiring Enzyme 1 Alpha (IRE1α) by High- 121. Dejeans, N.; Barroso, K.; Fernandez-Zapico, M. E.; Throughput Screening Using a DELFIA Assay. J. Biomol. et al. Novel Roles of the Unfolded Protein Response in the Screen. 2012, 18, 298–308. Control of Tumor Development and Aggressiveness. Semin. 138. Kimata, Y.; Ishiwata-Kimata, Y.; Ito, T.; et al. Two Cancer Biol. 2015, 33, 67–73. Regulatory Steps of ER-Stress Sensor Ire1 Involving Its 122. Sidrauski, C.; Acosta-Alvear, D.; Khoutorsky, A.; et al. Cluster Formation and Interaction with Unfolded Proteins. Pharmacological Brake-Release of mRNA Translation J. Cell Biol. 2007, 179, 75–86. Enhances Cognitive Memory. Elife 2013, 2, e00498. 139. Marcu, M. G.; Doyle, M.; Bertolotti, A.; et al. Heat Shock 123. Gallagher, C. M.; Walter, P. Ceapins Inhibit ATF6α Protein 90 Modulates the Unfolded Protein Response by Signaling by Selectively Preventing Transport of ATF6α to Stabilizing IRE1alpha. Mol. Cell Biol. 2002, 22, 8506–8513. the Golgi Apparatus during ER Stress. Elife 2016, 5, e11880. 140. Papa, F. R.; Zhang, C.; Shokat, K.; et al. Bypassing a Kinase 124. Rong, J.; Pass, I.; Diaz, P. W.; et al. Cell-Based High- Activity with an ATP-Competitive Drug. Science. 2003, Throughput Luciferase Reporter Gene Assays for Identifying 302, 1533–1537. and Profiling Chemical Modulators of Endoplasmic 141. Cerezo, M.; Lehraiki, A.; Millet, A.; et al. Compounds Reticulum Signaling Protein, IRE1. J. Biomol. Screen. 2015, Triggering ER Stress Exert Anti-Melanoma Effects and 20, 1232–1245. Overcome BRAF Inhibitor Resistance. Cancer Cell 2016, 125. Martinez, N. J.; Rai, G.; Yasgar, A.; et al. A High- 29, 805–819. Throughput Screen Identifies 2,9-Diazaspiro[5.5]Undecanes 142. Mimura, N.; Fulciniti, M.; Gorgun, G.; et al. Blockade of XBP1 as Inducers of the Endoplasmic Reticulum Stress Response Splicing by Inhibition of IRE1alpha Is a Promising Therapeutic with Cytotoxic Activity in 3D Glioma Cell Models. PLoS Option in Multiple Myeloma. Blood 2012, 119, 5772–5781. One 2016, 11, e0161486. 143. Obeng, E. A.; Carlson, L. M.; Gutman, D. M.; et al. Proteasome 126. Flaherty, D. P.; Miller, J. R.; Garshott, D. M.; et al. Discovery Inhibitors Induce a Terminal Unfolded Protein Response in of Sulfonamidebenzamides as Selective Apoptotic CHOP Multiple Myeloma Cells. Blood 2006, 107, 4907–4916. Pathway Activators of the Unfolded Protein Response. ACS 144. Yaw, D.; Dadey, A. Contributions of the Endoplasmic Med. Chem. Lett. 2014, 5, 1278–1283. Reticulum Stress Response to Glioblastoma Cell Viability. 127. Plate, L.; Cooley, C. B.; Chen, J. J.; et al. Small Molecule Ph.D. Thesis, Washington University in St. Louis, December Proteostasis Regulators That Reprogram the ER to Reduce 2015. Extracellular Protein Aggregation. Elife 2016, 5, e15550. 145. Waller, D. D.; Jansen, G.; Golizeh, M.; et al. A Covalent 128. Axten, J. M.; Romeril, S. P.; Shu, A.; et al. Discovery of Cysteine-Targeting Kinase Inhibitor of Ire1 Permits GSK2656157: An Optimized PERK Inhibitor Selected for Allosteric Control of Endoribonuclease Activity. Chem. Bio. Preclinical Development. ACS Med. Chem. Lett. 2013, 4, Chem. 2016, 17, 843–851. 964–968. 146. Tang, C. H.; Ranatunga, S.; Kriss, C. L.; et al. Inhibition of ER 129. Sanches, M.; Duffy, N. M.; Talukdar, M.; et al. Structure Stress-Associated IRE-1/XBP-1 Pathway Reduces Leukemic and Mechanism of Action of the Hydroxy Aryl Aldehyde Cell Survival. J. Clin. Invest. 2014, 124, 2585–2598. Class of IRE1 Endoribonuclease Inhibitors. Nat. Commun. 147. Dadey, D. Y. A.; Kapoor, V.; Khudanyan, A.; et al. The ATF6 2014, 5, 4202. Pathway of the ER Stress Response Contributes to Enhanced 130. Lhomond, S.; Pallares, N.; Barroso, K.; et al. Adaptation of Viability in Glioblastoma. Oncotarget 2016, 7, 2080–2092. the Secretory Pathway in Cancer through IRE1 Signaling. Methods Mol. Biol. 2015, 1292, 177–194.

68 Chapter 2: Conclusions and Contributions

In this second chapter we have gone through the UPR canonical and non-canonical signalling consolidating the information acquired in chapter 1. Moreover we have revisited the main players in UPR signalling, PERK, ATF6 and IRE1 and have begun our focus on IRE1 as the most evolutionarily conserved UPR mediator shaping the way to it becoming the centrepiece of this thesis. Taking a broad perspective of the available screening strategies and reporters of ER function and UPR activation we have summarised and analysed in vitro approaches to UPR modulator discovery and synthesis. Continuing on our journey from general ER functions, to the UPR, to drug discovery to shaping the hypothesis of this thesis; IRE1 is singled out as a pharmacological target in various disease models and we have briefly explored its association with GBM pathophysiology.

The author of this thesis contributed to the publication of “Control of the Unfolded Protein Response in Health and Disease” the following: Original text for a share of main text attributed to Chapters 2.1-2.3 and 2.5 inclusive, original text for the “Abstract” and “Concluding Remarks” submitted to the SLAS Discovery journal. Artistically he provided Figure 17. Finally he provided writing quality control, editorial duties and coordinated efforts to produce a homogeneous manuscript of publishable quality8.

7 Special thanks to Dr E. Chevet for his input in shaping figure 1. 8 Special thanks to Dr A. Avril, Dr N. Dejeans and Dr E. Chevet for their contribution in providing main text to chapter 2.4 and to shaping the final manuscript.

69 Chapter 3 Glioblastoma multiforme and IRE1 – clinical implications of translational research

70 Chapter 3: Foreword

71 Glioblastoma multiforme and IRE1 – clinical implications of translational research

History of GBM It has been 90 years since Glioblastoma multiforme was first established as a separate clinical entity by Harvey Cushing and Percival Bailey in their seminal work on histological glioma characterisation1. They pointed out in their manuscript that spongioblastoma multiforme as it was then known represented not only the largest group of glioma patients but also the one with the least favourable prognosis as when resection was attempted, it was observed that the tumour merged with the surrounding cerebral tissue making any demarcation impossible1. Whilst the works by Cushing and Bailey provided the basis for the modern classification of glioma, the early 1900s were not the earliest occurrences of glioblastoma in published literature. Burns in 1800 and Abernety in 1804 as well as Virchow in 1865 gave accounts with the latter providing the first histopathological description of GBM2 under the names telangiectatic or haemorrhagic glioma. The name glioblastoma multiforme was finally coined to avoid confusion with the benign polar spongioblastoma; a name too similar to the one used by Cushing and Bailey3. The importance of histopathological classification and the invasive and intracerebral metastatic properties of GBM were evident from reports of procedures involving complete hemispherectomies of the primary tumour site in patients who eventually succumbed to the disease from tumour recurrence despite the significant removal of brain tissue4. This notion of invasiveness was further characterised by the work of Scherer5 who described secondary tumour structures arising from healthy brain cellular compositions leading to perivascular and perineuronal lesions as opposed to primary sites of tumour growth such as around microvascular proliferative areas or pseudopalisading necroses6,7. Globally GBM in particular and CNS tumours in general have been in the sphere of clinical entity classification by the World Health Organisation since 1956 when a resolution of the WHO executive board preceded the World Health Assembly in 1957 to initiate the international classification of human tumours8. CNS tumours became part of the classification in 1979 when the first histological typing of tumours was published9. Subsequently another three editions followed reflecting the advances in immunohistochemistry leading to improved diagnostic pathology10, the genetic profiling in shaping CNS tumour classification and diagnosis11 and the familial improved classification of tumours along with predisposition

72 Figure 1. “GBM timeline”. Non-exhaustive timeline of significant events in the characterisation and treatment that shape the current clinical understanding and therapeutic options in GBM. characteristics8. The latest version of this classification was published in 2016 using molecular parameters as well as histology, providing guidelines of how tumour diagnoses should be approached. This version did indeed impact the consensus classification of GBM as well, incorporating IDH-wildtype and IDH mutant strains of the tumour12. As such the clinical entity of GBM has been through a 200 year journey (Figure 1) of evolution, recognition and therapeutic development, making the fact that it remains a wholly fatal diagnosis the more confirmatory of the difficulties confronted by physicians and scientists alike when attempting to treat it.

Diagnosis, Prognosis and current therapeutic interventions GBM is the commonest primary CNS tumour with an incidence of 3 per 10000013. In the United States in 2017 alone there were roughly 80000 new predicted diagnoses of CNS neuroepithelial malignancies 50% of which were expected to be GBMs14. The disease has a dismal prognosis with patients succumbing to the tumour between 15 and 18 months post diagnosis, with a median 5 year survival at less than 6%15. These outcomes are the result of a highly necrotic and endothelial proliferative histopathological picture that merits the highest and most malignant grade in the WHO classification of brain tumours (grade IV). GBMs account for more than half of all glioma diagnoses and can be primary or secondary. Primary GBMs are arising spontaneously and conferring a worse prognosis than their secondary paralogues. Secondary GBMs arise from the progression of grade II or grade III gliomas that may be either astrocytomas or oligodendrogliomas in nature16. The most recent comprehensive longitudinal analyses of the

73 available GBM transcriptomes categorises GBMs into three distinct subtypes, termed proneural, mesenchymal and classical17. The mesenchymal phenotype seems to confer a worse prognosis than the classical and more so than the proneural phenotype however, there is a prevalent intra- tumoral plasticity that may alter any given GBM tumour phenotype according to its microenvironment18. A distinctive characteristic of secondary GBMs is the prevalence of mutations in the NADP+ dependent isocitrate dehydrogenase genes (IDH1 and 2) and although the clinical presentations of end stage primary and secondary GBMs are indistinguishable, mutations in the IDH genes confer better prognostic characteristics than those without these mutations. New variants of GBM have recently joined the prognostically worse IDH WT tumours in the form of epithelioid GBM, giant cell GBM and gliosarcoma with epihelioid GBMs having a propensity to affect children and younger adults12. A subgroup of GBMs is part of the latest WHO classification termed NOS for which IDH evaluation cannot be performed based on the combination of IDH sequencing and R132H IDH1 biopsy immunohistochemistry for different patient age groups19,20. Other markers of GBM progression are routinely used in clinical practice and they include MGMT promoter methylation, EGFRvIII amplification and 1p19q deletion21. There is a certain dichotomy between primary or secondary GBMs when it comes to the mutational burden that may characterise them. Primary tumours are characterised by EGFR amplification, NF1, RB1, PTEN and TERT promoter mutations and CDKN2A-p16INK4a deletions22 whilst secondary tumours present with TP53 mutations, MGMT promotor methylation23 and IDH1 mutations whilst loss of RB1 or CDKN2A may cause low grade gliomas to progress to secondary GBMs. Missense mutations in the chromatin regulator ATRX are mutually exclusive with 1p/19q co-deletions and are hallmarks of oligodendroglioma development however, such mutations closely correlate with IDH1/2 and TP53 mutations that pertain to secondary GBM; as such, it is clear that there is a certain degree of inter-tumoral plasticity providing further evidence of the ability of grade II/III gliomas to progress to GBM24. Epigenetic silencing of the O6-Methylguanine-DNA-methyltransferase (MGMT), a protein that removes alkyl groups from DNA and is important in DNA repair, has been associated with better prognosis as tumours are less likely to be able to repair damage cause by alkylating agents such as TMZ, with GBM patients presenting with methylated MGMT promoter GBM tumours benefiting more from TMZ treatment compared to those who did not possess the methylated promoter25. These highly aggressive tumours are very rare in childhood with less than 10% of paediatric CNS tumours being GBMs26. The heterogeneous nature of the disease and the speed of its progression have made any public health analysis of population based characteristics difficult however, it seems that the average age of onset in primary GBMs is higher (62 years) than that of secondary GBMs (45 years) albeit the age of presentation of the grade II or III gliomas that secondary GBMs derive from is a lot more diverse27. Men seem to be more

74 frequently affected than women at a 3:1 ratio but for both populations the survival rapidly decreases in the second year post diagnosis22. Other than the presence of IDH mutations pointing towards the primary/secondary dichotomy the only other two factors that are useful in determining prognosis are age (better with young age) and a good Karnofsky Performance Status (KPS), a scale designed to assess a patient’s ability to be independent and their dependence on clinical assistance28. Albeit highly invasive and without ability to metastasise outside the CNS there is no clear anatomical effect on the outcomes of tumour progression however, the positioning of the tumour plays a significant indirect prognostic role. Prognosis is inextricably linked to the tumour operability and to maximisation of resection. If the tumour is invading vital structures or if it is deep enough for the surgery to be potentially breaching considerations of non- maleficence then survival is rapidly decreased. Studies have attempted to classify the anatomical incidence of gliomas since the 1940s with many recent investigations shedding some light on the potential anatomical incidence and molecular phenotype correlation of GBMs. As can be seen in figure 2, although this is by no means an exhaustive guide of GBM presentation, the majority of GBMs seem to develop in the frontal lobes with temporal lobes and a combination of parietal lobes and deeper structures vying for second place. Moreover in the Finnish study of 331 patients where these results were observed, there was an increased tumoral presentation in the right hemisphere compared to the left29. Radiological investigations using magnetic resonance data acquired at diagnosis and during operative consultations showed in vivo that

Figure 2. “Anatomical GBM incidence”. Graphical representations of all three brain section planes. These are meant to be representations only, any specific anatomical structures that can be discerned in the diagrams are not to be considered as key to the interpretation of GBM presentation. The middle coronal section is seen from the front and in the bottom sagittal section the right hemisphere is visible. Data that provided information for this figure can be found in Larjavaara et al29.

75 glioma cells preferentially migrate along white matter fibres to invade neighbouring structures30. Although inter-tumour variability has proven difficult to categorise and annotate, efforts have been made to generate intra-tumoral anatomical and molecular maps utilising genomic changes in GBM documented in repositories such as The Cancer Genome Atlas31 (TCGA) and the Repository for Molecular Brain Neoplasia Data32 (REMBRANDT) to pinpoint the role of such alterations in the pathophysiology of the disease. RNA sequencing, in situ hybridisation and laser microdissection were employed to scrutinise tumoral anatomic features to produce the Ivy Glioblastoma Atlas Project (Ivy GAP)33. Further genetic investigations of brain cancers led to the development of web portals such as GlioVis where molecular profiling of GBMs is a readily available service34. As such it is evident that research into the details of GBM pathophysiology is of great interest with multiple public online tools providing access to a wealth of information on the disease on a molecular and anatomical level. Despite this wealth of information available to researchers the outcome for patients has not improved. Currently there are multiple investigations into the alternative therapeutic avenues that can be exploited in GBM. However, patients can currently only rely on a therapeutic regime of a combination of surgery, chemotherapy and radiotherapy. Surgery has to be maximal but safe considering the infiltrative nature of the tumour which pertains to an increased removal of healthy brain tissue along with malignant one. Resection does not only minimise tumour mass and thus alleviate symptoms conferred by space occupying lesion characteristics but also provide brain tissue for histological analysis to predict prognosis and hence guide post-operative and pharmacological management35. Indeed pre- or intra-surgical planning to spare healthy tissue has been the focus of a few studies with brain dynamics approaches such as the Virtual Brain36 and intra-operative approaches such as rapid evaporative ionisation mass spectrometry (REIMS) coupled with the electrosurgical iKnife37 designed to address the issue of sparing valuable brain tissue. As discussed earlier in this chapter good prognosis is certainly dependent on a patient’s KPS. Having established that it is possible that the majority of tumours do impact the frontal and temporal lobes, GBMs affect cognition and behaviour thus heavily reducing a patient’s KPS. Radiotherapy is a proven addition to GBM and although treatment optimisation is very much dependent on how old or frail patient are, on the state of disease recurrence and the available imaging techniques such as fMRIs to determine accurate dosing, there are significant considerations for radiotherapy to be used a s an adjuvant treatment alongside immunotherapy and targeted therapy against cancer stem cells by targeting GBM stem cell niches at their reported source at the Sub Ventricular Zone (SVZ)38. Targeting the SVZ is also supported by findings in patient brain sections and murine GBM models that point to neural stem cells containing GBM driver mutations (such as TP53, ERT and KEL) which

76 migrate from the SVZ to lead to the development of high-grade gliomas like GBM in multiple brain regions. This suggests that the SVZ may not only be the source of chemo-resistance through GBM stem cell properties39 but also the origin of certain high grade gliomas40. The chemotherapeutic of choice in GBM is temozolomide (TMZ) whose conception can be traced back to the 1970s and heterocycle chemical research based on studies on the tumour inhibitory effects of triazenes41. Such molecules were the backbone of a group of tetrazine alkylating agents of the family of imidazotetrazinone molecules, the most promising candidate of which, mitozolomide, fared badly in phase I clinical trials as it displayed extreme thrombocytopaenic toxicity42. These were the precursors of CCRG81045, a triazene presented as a an alternative to dacarbazine43 which eventually became TMZ, a more benign form of mitozolomide due to its lack of ability to crosslink with DNA resulting from a chloride to methyl tail substitution change. The first phase I clinical trial of TMZ and subsequent investigations in dose escalations over five days showed positive results particularly in patients suffering from gliomas and hence TMZ was brought about as a preferred chemotherapeutic agent in GBM44,45. This current form of consensus GBM treatment was the result of a pilot phase II trial investigating concomitant administration of TMZ with fractionated radiotherapy, preceding a further 6 rounds of TMZ treatment showing a 31% 2 year survival46 and a subsequent phase III trial commissioned by the European Organisation for Research and Treatment of Cancer (EORTC), Brain Tumour and Radiotherapy Groups and the National Cancer Institute of Canada (NCIC) resulting in a 26.5% 2 year survival rate47. It is evident that a multidisciplinary approach to clinical problem solving has brought advances to the field of GBM research with medicinal chemistry and clinical medicine at the forefront of drug discovery. Incorporating basic functions of cancer biology into this landscape is thus enriching the possibilities for targeted research and thus therapy.

Current Clinical Trials in GBM Reflecting the importance of the GBM burden on public health and the difficulty to combat it, there are a plethora of clinical trials currently conducted on GBM. A meta-analysis of clinical trials in GBM between 2005 and 2017 pinpointed more than 400 clinical trials carried out exclusively on GBM, the majority of which were phase IIs48. The aggressiveness of GBM combined with low patient participation and long clinical trial development times do impede successful trial completion and meaningful result acquisition49. Moreover biological barriers to success such as the impermeability of the blood brain barrier50 and the toxicity of chemotherapeutics, radiotherapy and surgery, the most common result of which being cerebral oedema51, make CNS drug discovery more challenging. However, there are currently at least 10

77 phase III and 2 phase IV trials that explore novel therapies in GBM that span immunotherapy, gene therapy, sensitisation adjuvant therapies or targeted therapies. A promising phase IV trial is looking into Optune therapy that although does not improve survival reports on markedly improved quality of life for GBM patients52. VB-111 is an adenoviral gene therapy53 was investigated as an anti-neovascularisation agent in the presence of the vascular endothelial growth factor bevazicumab (GLOBAL) in GBM54 without showing improvement when compared to avastin alone. Purified retroviral Toca 511 and flucytosine based Toca FC are also tested investigating the response of patients with recurrent disease55. Nivolumab, an immune checkpoint blockade agent, has also been studied through trials CheckMate 14356 and 49857 comparing it to bevazicumab and temozolomide respectively. CART cells peripherally infused and EGFRvIII-directed showed no off tumour toxicity or cytokine release syndrome and hence promise in GBM treatment58. Additionally immunotherapy dendritic cell vaccines have shown promise through DCVax-L59 as well as dendritic cells pulsed with immunogenic antigens through ICT-10760. Drug repurposing has also been explored with disulfiram, a common generic used to treat alcoholism, proposed to affect GBM stem like cells in vitro and in vivo61 and forwarded into a phase III62 trial to test its efficacy along with dietary copper a supplement. Furthermore the STELLAR study investigates the kinase inhibitor sunitinib as an antiangiogenic and gliomagenesis inhibitor63. Collectively these studies show that different approaches are being investigated in the search for a solution to GBM with more being tested pre-clinically, to bring therapies successful in other diseases to the GBM field, an example of which being engineering CAR T cells with adhesion molecules (ALCAM) to cross the endothelial barrier of a murine model of GBM64,65. Despite persistent and numerous attempts to find the key to solving the GBM problematic though, there are still no effective therapeutic solutions available66. Therefore the search for new targets and therapeutic approaches is ever more pertinent to the patients who currently are faced with or will face this grim diagnosis.

IRE1 involvement in GBM pathophysiology The current landscape of clinical trials in GBM offers a certain degree of rationality in targeting IRE1 in GBM. As we saw above, studies are focusing on drug repurposing and on targeting the stem phenotype of GBM cells with simple drugs such as disulfiram; such investigations would support exploring IRE1’s suitability as a target due to its ubiquitous nature and its documented involvement in stemness67. Moreover, therapeutics in GBM trials such as sunitinib (STELLAR)

78 Figure 3. “Studying IRE1 in GBM”. A chronological approach of studies that have explored the involvement of IRE1 in GBM to date and have shown its effect in aspects of its pathophysiology and pathogenesis. From top to bottom it documents attempts t understand the process that guides GBM from cell (top) to tumour formation and maintenance (bottom).

were shown to affect IRE1’s activity68,69. These indications collectively make an investigation into targeting IRE1 in GBM all the more attractive. As briefly introduced in chapter 2, IRE1 has been shown to be an integral part of GBM pathogenesis and maintenance and this has been the culmination of more than 10 years of research into the particular attributes of IRE1 involvement specifically in GBM biology. IRE1 was shown to play a role in the ischaemic response to hypoxia and hypoglycaemia in GBM cells by upregulating vascular endothelial factor-A (VEGF-A)70. On top of angiogenesis it was demonstrated that IRE1 is a key regulator of malignant glioma progression as tumours

79 expressing low levels of IRE1 showed reduced blood perfusion and growth rate71. GBM cells survive a heightened stress state, one that would signal normal physiological state cells for apoptosis, by adapting to a hostile microenvironment and IRE1 is a major regulator of this adaptation by controlling stress fibre formation and the regulation of extracellular matrix proteins such as SPARC, impacting tumour growth, infiltration and invasion72. Targets of the IRE1 endoribonuclease activity such as SPARC are among a list of factors impacting oncogenesis and indeed, PER1, another IRE1 substrate is responsible for tumour cell circadian clock components, underlying tumour development73. Moreover such mechanisms of cancer progression as GBM neovascularisation and invasion were shown to be part of the IRE1 activity repertoire74 whilst it was shown that as epidermal growth factor receptors (EGFR) contribute to the development of malignant gliomas, epiregulin, an EGFR ligand, under the control of IRE1 is compounding this effect75. This timeline of IRE1 research in GBM can be seen in figure 3. Components of the IRE1 signalling cascade both upstream and downstream of IRE1 have also been shown to have an effect in GBM. IRE1 activity is under the direct control of the ER chaperone BiP, which has been shown to be significantly upregulated in malignant gliomas as opposed to normal brain parenchyma and its targeting mediated a sensitisation of glioma cells to TMZ76. This also occurred in studies with BiP targeting antibodies exhibiting anti-tumour activity which sensitised tumour cells to radiotherapy77. Moreover, proteasomal inhibition with bortezomib compounded GBM cytotoxicity purportedly by elevating BiP and hence ER stress to levels beyond those that cells could cope with78. BMI1, a polycomb-group protein, shown to be intricately linked to IRE1 expression79, was discovered through a combination of ChIP-seq with in vivo RNAi screening to have specific targets important in oncogenesis39. XBP1, the major downstream target of IRE1, is linked to the Wnt pathway80 which in turn has been proposed to drive GBM stem cell differentiation and promote invasion81. IRE1 is responsible for the degradation and regulation of a group of microRNAs, molecules that show an increasing attractiveness as therapeutic targets and molecular mediators in GBM by regulating cell cycle events82. Finally mesenchymal migration has been proposed as a therapeutic target in GBM through the use of actin cytoskeleton and adhesion signalling molecules83, a process that has been extensively explored as a connector to IRE1 activity84 along with the epithelial mesenchymal transition (EMT) activation85, another pathway targeted in GBM through the EIS regiment86. IRE1 signalling is part of the wider ER stress response which is triggered upon perturbation of calcium homeostasis. Calcium homeostasis is an integral cog in the maintenance of GBM whole cell homeostasis compared to normal astrocytes as inhibition of the sarcoplasmic reticulum calcium metabolism pathways with the known ER stressor thapsigargin shows a significant

80 difference in pH regulation and calcium homeostasis in GBM cells further documenting the role of ER stress in the adaptation of GBM cells to changes in the microenvironment87. It is thus evident that although not well characterised, the involvement of IRE1 in GBM pathophysiology is a topic worth pursuing in translational research in order to discern and dissect further the exact role it may play in this fatal disease. Such enrichment of knowledge as to the physiological aspects of IRE1 manipulation in tumour cell homeostasis will be of paramount importance not only in the development of novel therapeutic approaches in GBM but also in the reinforcement of pre-clinical data that drive patient stratification and pharmacovigilance throughout the process of novel drug discovery or drug repurposing.

References 1. Bailey P, Cushing H. A Classification of the Tumors of the Glioma Group on a Histogenic Basis with a Correlated Study of Prognosis. Philadelphia: Lippincott; 1928. 2. Stoyanov G St., Dzhenkov DL. On the Concepts and History of Glioblastoma Multiforme - Morphology, Genetics and Epigenetics. Folia Med (Plovdiv) 2018;60(1):48–66. 3. Zülgh KJ. Biology and Morphology of Glioblastoma Multiforme. Acta Radiol Ther Phys Biol 1969;8(1–2):65–77. 4. Gardner WJ, Karnosh LJ, McClure Christopher C. JR, Gardner ANNK. Residual function following hemispherectomy for tumour and for infantile hemiplegia. Brain 1955;78(4):487–502. 5. Scherer HJ. The Pathology of Cerebral Gliomas. J Neurol Psychiatry 1940;3(2):147 LP-177. 6. Scherer HJ. Structural Development in Gliomas. Am J Cancer 1938;34(3):333 LP-351. 7. Scherer HJ. The forms of growth in gliomas and their practical significance. Brain 1940;63(1):1–35. 8. Louis DN, Ohgaki H, Wiestler OD, et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 2007;114(2):97–109. 9. Zülch KJ (Klaus J. Histological typing of tumours of the central nervous system. Geneva: World Health Organization; 1979. 10. Kleihues P, Burger PC, Scheithauer BW. Histological Classification of Tumours of the Central Nervous System. In: Histological Typing of Tumours of the Central Nervous System. Berlin, Heidelberg: Springer Berlin Heidelberg; 1993. p. 5–10. 11. Kleihues P, Louis DN, Scheithauer BW, et al. The WHO classification of tumors of the nervous system. J Neuropathol Exp Neurol 2002;61(3):215-25; discussion 226-9. 12. Louis DN, Perry A, Reifenberger G, et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol

81 2016;131(6):803–20. 13. Tamimi AF, Juweid M. Epidemiology and Outcome of Glioblastoma. Codon Publications; 2017. 14. Ostrom QT, Gittleman H, Xu J, et al. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2009–2013. Neuro Oncol 2016;18(suppl_5):v1–75. 15. Wirsching H-G, Galanis E, Weller M. Glioblastoma. In: Handbook of clinical neurology. 2016. p. 381–97. 16. Obacz J, Avril T, Le Reste P-J, et al. Endoplasmic reticulum proteostasis in glioblastoma- From molecular mechanisms to therapeutic perspectives. Sci Signal 2017;10(470):eaal2323. 17. Sidaway P. CNS cancer: Glioblastoma subtypes revisited. Nat Rev Clin Oncol 2017;14(10):587–587. 18. Wang Q, Hu B, Hu X, et al. Tumor evolution of glioma intrinsic gene expression subtype associates with immunological changes in the microenvironment. Cancer Cell 2017;32(1):42–56.e6. 19. Chen L, Voronovich Z, Clark K, et al. Predicting the likelihood of an isocitrate dehydrogenase 1 or 2 mutation in diagnoses of infiltrative glioma. Neuro Oncol 2014;16(11):1478–83. 20. Pisapia DJ. The Updated World Health Organization Glioma Classification: Cellular and Molecular Origins of Adult Infiltrating Gliomas. Arch Pathol Lab Med 2017;141(12):1633– 45. 21. Szopa W, Burley TA, Kramer-Marek G, Kaspera W. Diagnostic and Therapeutic Biomarkers in Glioblastoma: Current Status and Future Perspectives. Biomed Res Int 2017;2017:8013575. 22. Ohgaki H, Dessen P, Jourde B, et al. Genetic Pathways to Glioblastoma. Cancer Res 2004;64(19):6892–9. 23. Nakamura M, Watanabe T, Yonekawa Y, Kleihues P, Ohgaki H. Promoter methylation of the DNA repair gene MGMT in astrocytomas is frequently associated with G:C -- A:T mutations of the TP53 tumor suppressor gene. Carcinogenesis 2001;22(10):1715–9. 24. Reuss DE, Sahm F, Schrimpf D, et al. ATRX and IDH1-R132H immunohistochemistry with subsequent copy number analysis and IDH sequencing as a basis for an “integrated” diagnostic approach for adult astrocytoma, oligodendroglioma and glioblastoma. Acta Neuropathol 2015;129(1):133–46. 25. Hegi ME, Diserens A-C, Gorlia T, et al. MGMT Gene Silencing and Benefit from Temozolomide in Glioblastoma. N Engl J Med 2005;352(10):997–1003.

82 26. Das KK, Kumar R. Pediatric Glioblastoma. Codon Publications; 2017. 27. Ohgaki H. Genetic pathways to glioblastomas. Neuropathology 2005;25(1):1–7. 28. Adamson C, Kanu OO, Mehta AI, et al. Glioblastoma multiforme: a review of where we have been and where we are going. Expert Opin Investig Drugs 2009;18(8):1061–83. 29. Larjavaara S, Mäntylä R, Salminen T, et al. Incidence of gliomas by anatomic location. Neuro Oncol 2007;9(3):319–25. 30. Esmaeili M, Stensjøen AL, Berntsen EM, Solheim O, Reinertsen I. The Direction of Tumour Growth in Glioblastoma Patients. Sci Rep 2018;8(1):1199. 31. Cancer Genome Atlas Research Network TCGA (TCGA) R. Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature 2008;455(7216):1061–8. 32. Madhavan S, Zenklusen J-C, Kotliarov Y, Sahni H, Fine HA, Buetow K. Rembrandt: helping personalized medicine become a reality through integrative translational research. Mol Cancer Res 2009;7(2):157–67. 33. Puchalski RB, Shah N, Miller J, et al. An anatomic transcriptional atlas of human glioblastoma. Science 2018;360(6389):660–3. 34. Bowman RL, Wang Q, Carro A, Verhaak RGW, Squatrito M. GlioVis data portal for visualization and analysis of brain tumor expression datasets. Neuro Oncol 2017;19(1):139– 41. 35. Manrique-Guzmán S, Herrada-Pineda T, Revilla-Pacheco F. Surgical Management of Glioblastoma. Codon Publications; 2017. 36. Aerts H, Schirner M, Jeurissen B, et al. Modeling Brain Dynamics in Brain Tumor Patients Using the Virtual Brain. eNeuro 2018;5(3):ENEURO.0083-18.2018. 37. Balog J, Sasi-Szabó L, Kinross J, et al. Intraoperative tissue identification using rapid evaporative ionization mass spectrometry. Sci Transl Med 2013;5(194):194ra93. 38. Mann J, Ramakrishna R, Magge R, Wernicke AG. Advances in radiotherapy for glioblastoma. Front. Neurol. 2018;8(JAN):748. 39. Gargiulo G, Cesaroni M, Serresi M, et al. In Vivo RNAi Screen for BMI1 Targets Identifies TGF-ɅȀ-ER Stress Pathways as Key Regulators of Neural- and Malignant Glioma- Stem Cell Homeostasis. Cancer Cell 2013;23(5):660–76. 40. Lee JH, Lee JE, Kahng JY, et al. Human glioblastoma arises from subventricular zone cells with low-level driver mutations. Nature 2018;560(7717):243–7. 41. Audette RCS, Connors TA, Mandel HG, Merai K, Ross WCJ. Studies on the mechanism of action of the tumour inhibitory triazenes. Biochem Pharmacol 1973;22(15):1855–64. 42. Newlands ES, Blackledge G, Slack JA, et al. Phase I clinical trial of mitozolomide. Cancer Treat Rep 69(7–8):801–5.

83 43. Stevens MF, Hickman JA, Langdon SP, et al. Antitumor activity and pharmacokinetics in mice of 8-carbamoyl-3-methyl-imidazo[5,1-d]-1,2,3,5-tetrazin-4(3H)-one (CCRG 81045; M & B 39831), a novel drug with potential as an alternative to dacarbazine. Cancer Res 1987;47(22):5846–52. 44. Brampton MH, S NE, R BG, et al. Phase I trial of temozolomide (CCRG 81045: M&B 39831: NSC 362856). Br J Cancer 45. Bower M, Newlands ES, Bleehen NM, et al. Multicentre CRC phase II trial of temozolomide in recurrent or progressive high-grade glioma. Cancer Chemother Pharmacol 1997;40(6):484–8. 46. Stupp R, Dietrich P-Y, Ostermann Kraljevic S, et al. Promising survival for patients with newly diagnosed glioblastoma multiforme treated with concomitant radiation plus temozolomide followed by adjuvant temozolomide. J Clin Oncol 2002;20(5):1375–82. 47. Stupp R, Mason WP, van den Bent MJ, et al. Radiotherapy plus Concomitant and Adjuvant Temozolomide for Glioblastoma. N Engl J Med 2005;352(10):987–96. 48. Vanderbeek AM, Rahman R, Fell G, et al. The clinical trials landscape for glioblastoma: is it adequate to develop new treatments? Neuro Oncol 2018;20(8):1034–43. 49. Butlen-Ducuing F, Pétavy F, Guizzaro L, et al. Regulatory watch: Challenges in drug development for central nervous system disorders: a European Medicines Agency perspective. Nat Rev Drug Discov 2016 1512 2016; 50. Daneman R, Prat A. The blood-brain barrier. Cold Spring Harb Perspect Biol 2015;7(1):a020412. 51. Cohen J V, Kluger HM. Systemic Immunotherapy for the Treatment of Brain Metastases. Front Oncol 2016;6:49. 52. Topfer L-A, Farrah K. Alternating Electric Fields (“Tumour-Treating Fields”) for the Treatment of Glioblastoma. Canadian Agency for Drugs and Technologies in Health; 2016. 53. Gruslova A, Cavazos DA, Miller JR, et al. VB-111: a novel anti-vascular therapeutic for glioblastoma multiforme. J Neurooncol 2015;124(3):365–72. 54. A Phase 3, Pivotal Trial of VB-111 Plus Bevacizumab vs. Bevacizumab in Patients With Recurrent Glioblastoma (GLOBE) - Full Text View - ClinicalTrials.gov [Internet]. [cited 2019 Jan 20];Available from: https://clinicaltrials.gov/ct2/show/NCT02511405?term=glioblastoma&phase=2 55. Cloughesy TF, Landolfi J, Vogelbaum MA, et al. Durable complete responses in some recurrent high-grade glioma patients treated with Toca 511 + Toca FC. Neuro Oncol 2018;20(10):1383–92. 56. A Study of the Effectiveness and Safety of Nivolumab Compared to Bevacizumab and of

84 Nivolumab With or Without Ipilimumab in Glioblastoma Patients - Full Text View - ClinicalTrials.gov [Internet]. [cited 2019 Jan 20];Available from: https://clinicaltrials.gov/ct2/show/NCT02017717?term=glioblastoma&phase=2 57. An Investigational Immuno-therapy Study of Nivolumab Compared to Temozolomide, Each Given With Radiation Therapy, for Newly-diagnosed Patients With Glioblastoma (GBM, a Malignant Brain Cancer) - Full Text View - ClinicalTrials.gov [Internet]. [cited 2019 Jan 20];Available from: https://clinicaltrials.gov/ct2/show/NCT02617589?term=glioma%2C+glioblastoma%2C+ brain+tumor&phase=2 58. O’Rourke DM, Nasrallah MP, Desai A, et al. A single dose of peripherally infused EGFRvIII-directed CAR T cells mediates antigen loss and induces adaptive resistance in patients with recurrent glioblastoma. Sci Transl Med 2017;9(399):eaaa0984. 59. Liau LM, Ashkan K, Tran DD, et al. First results on survival from a large Phase 3 clinical trial of an autologous dendritic cell vaccine in newly diagnosed glioblastoma. J Transl Med 2018;16(1):142. 60. A Study of ICT-107 Immunotherapy in Glioblastoma Multiforme (GBM) - Full Text View - ClinicalTrials.gov [Internet]. [cited 2019 Jan 20];Available from: https://clinicaltrials.gov/ct2/show/NCT01280552 61. Azar K, Kannappan V, Liu Y, et al. Disulfiram targets glioblastoma-stem-like cells in vitro and in vivo. Neuro Oncol 2018;20(suppl_1):i20–i20. 62. Jakola AS, Werlenius K, Mudaisi M, et al. Disulfiram repurposing combined with nutritional copper supplement as add-on to chemotherapy in recurrent glioblastoma (DIRECT): Study protocol for a randomized controlled trial. F1000Research 2018;7:1797. 63. A Phase II/III Study of High-dose, Intermittent Sunitinib in Patients With Recurrent Glioblastoma Multiforme - Full Text View - ClinicalTrials.gov [Internet]. [cited 2019 Jan 20];Available from: https://clinicaltrials.gov/ct2/show/NCT03025893?term=glioblastoma&phase=2 64. Brown MH, Dustin ML. Steering CAR T Cells into Solid Tumors. N Engl J Med 2019;380(3):289–91. 65. Samaha H, Pignata A, Fousek K, et al. A homing system targets therapeutic T cells to brain cancer. Nature 2018;561(7723):331–7. 66. Le Reste P-J, Avril T, Quillien V, Morandi X, Chevet E. Signaling the Unfolded Protein Response in primary brain cancers. Brain Res 2016;1642:59–69. 67. ÀǡÒǡ®ǡǤ stress in stemness, pluripotency and development. Eur. J. Cell Biol. 2016;95(3–5):115–23. 68. ǡ ǡǡǤ ͝Ʉ

85 Endoribonuclease Outputs to Determine Divergent Cell Fates. Cell 2009;138(3):562–75. 69. Korennykh A V, Egea PF, Korostelev AA, et al. The unfolded protein response signals through high-order assembly of Ire1. Nature 2009;457(7230):687–93. 70. Drogat B, Auguste P, Nguyen DT, et al. IRE1 signaling is essential for ischemia-induced vascular endothelial growth factor-A expression and contributes to angiogenesis and tumor growth in vivo. Cancer Res 2007;67(14):6700–7. 71. Auf G, Jabouille A, Guérit S, et al. Inositol-requiring enzyme 1alpha is a key regulator of angiogenesis and invasion in malignant glioma. Proc Natl Acad Sci U S A 2010;107(35):15553–8. 72. Dejeans N, Pluquet O, Lhomond S, et al. Autocrine control of glioma cells adhesion and migration through IRE1alpha-mediated cleavage of SPARC mRNA. J Cell Sci 2012;125(Pt 18):4278–87. 73. Pluquet O, Dejeans N, Bouchecareilh M, et al. Posttranscriptional regulation of PER1 underlies the oncogenic function of IREalpha. Cancer Res 2013;73(15):4732–43. 74. Jabouille A, Delugin M, Pineau R, et al. Glioblastoma invasion and cooption depend on ͝Ʉ Ǥ ͜͞͝͡Ǣ͢ȋͤ͞Ȍǣͥ͞͠͞͞–34. 75. Auf G, Jabouille A, Delugin M, et al. High epiregulin expression in human U87 glioma ͝Ʉ Ǥ 2013;13:597. 76. Pyrko P, Schonthal AH, Hofman FM, Chen TC, Lee AS. The Unfolded Protein Response Regulator GRP78/BiP as a Novel Target for Increasing Chemosensitivity in Malignant Gliomas. Cancer Res 2007;67(20):9809–16. 77. Dadey DYA, Kapoor V, Hoye K, et al. Antibody targeting GRP78 enhances the efficacy of radiation therapy in human glioblastoma and non-small-cell lung cancer cell lines and tumor models. Clin Cancer Res 2016; 78. Kardosh A, Golden EB, Pyrko P, et al. Aggravated Endoplasmic Reticulum Stress as a Basis for Enhanced Glioblastoma Cell Killing by Bortezomib in Combination with Celecoxib or Its Non-Coxib Analogue, 2,5-Dimethyl-Celecoxib. Cancer Res 2008;68(3):843–51. 79. Li X-X, Zhang H-S, Xu Y-M, et al. ͝Ʉ sis Ʌ- ͝Ʉ Ǥ Oncogene 2017;36(48):6738–46. 80. Cho YM, Kim DH, Kwak S-N, Jeong S-W, Kwon O-J. X-box binding protein 1 enhances adipogenic differentiation of 3T3-L1 cells through the downregulation of Wnt10b expression. FEBS Lett 2013;587(11):1644–9. 81. Hu B, Wang Q, Wang YA, et al. Epigenetic Activation of WNT5A Drives Glioblastoma Stem Cell Differentiation and Invasive Growth. Cell 2016;167(5):1281–1295.e18.

86 82. Teplyuk NM, Uhlmann EJ, Gabriely G, et al. Therapeutic potential of targeting microRNA-10b in established intracranial glioblastoma: first steps toward the clinic. EMBO Mol Med 2016;8(3):268–87. 83. Zhong J, Paul A, Kellie SJ, O’Neill GM. Mesenchymal migration as a therapeutic target in glioblastoma. J Oncol 2010;2010:430142. 84. Urra H, Henriquez DR, Cánovas J, et al. ͝Ʉ migration through a direct interaction with filamin A. Nat Cell Biol 2018;20(8):942–53. 85. Cuevas EP, Eraso P, Mazón MJ, et al. LOXL2 drives epithelial-mesenchymal transition via activation of IRE1-XBP1 signalling pathway. Sci Rep 2017;7(1):44988. 86. Kast RE, Skuli N, Karpel-Massler G, Frosina G, Ryken T, Halatsch M-E. Blocking epithelial-to-mesenchymal transition in glioblastoma with a sextet of repurposed drugs: the EIS regimen. Oncotarget 2017;8(37):60727–49. 87. Kovacs GG, Zsembery A, Anderson SJ, et al. Changes in intracellular Ca2+ and pH in response to thapsigargin in human glioblastoma cells and normal astrocytes. Am J Physiol Cell Physiol 2005;289(2):C361-71.

87 Chapter 3: Conclusions

In this third and final chapter of the introduction we have gone through the clinical concept of GBM and have established why it merits the attention that it is receiving as a clinical entity from the clinical and the translational scientific community. It is a dismal disease with much still to be discovered as to its origins and progression. After having outlined the history that has shaped the modern classification of GBM, the current therapeutic options to patients were investigated. Although it is evident that GBM research benefits from the full complement of available novel techniques and research methods it is still in dire need for further characterisation at the molecular and macroscopic level in order to achieve true patient stratification and hence improve upon treatment outcomes. Upon summarising the current promising clinical trials in GBM, focusing on phase III trials, it was evident that themes such as stemness, differentiation, drug repurposing, small molecule therapeutics, tumour sensitisation to current therapies through adjuvant treatments and need for patient stratification are of great interest in the development of a detailed landscape of GBM pathophysiology, diagnosis, prognosis and treatment planning. Such themes are very much in the sphere of control of the UPR and IRE1 in particular, something we have explored in the latter part of this chapter where despite the unassailable position of the UPR as a central signalling cascade in GBM, we put IRE1 under the microscope to discern is role in GBM so far. This has yielded information that pertains to the fact that although there are strong indicators of IRE1 involvement in GBM, further characterisation is paramount to linking the above concepts with IRE1 activity. If such link was established then IRE1 would undeniably become a target of pharmacological interest in GBM.

88 Hypothesis & Objectives

Throughout the first 3 introductory chapters of this thesis we have established that Endoplasmic Reticulum signalling in general and the Unfolded Protein Response in particular play a significant role in health and disease. Moreover the UPR transducer IRE1 plays an important role in the realisation of the UPR and has been shown to be involved in the pathophysiology of GBM, a wholly fatal diagnosis in dire need for novel therapeutic options. As such the following hypothesis statement is hereafter proposed to be tested:

“IRE1 is a major pathophysiological mediator and valid pharmacological target in GBM and its modulation may provide novel therapeutic options as an adjuvant disease modifying

treatment.”

This hypothesis will be tested via three major primary objectives:

1. Generate a translational cellular model of altered IRE1 activity in GBM. 2. Characterise the impact of IRE1 signalling on GBM pathophysiology and treatment resistance. 3. Generate novel IRE1 modulators and test their effect on GBM sensitivity to temozolomide; the current chemotherapeutic option in GBM.

89 Results

Having acquired all the pertinent information through literature research to form a testable hypothesis we will set out in the next three chapters to explore how each of the three major objectives has been approached and met. To demonstrate how a cellular translational model of altered IRE1 activity in GBM was designed and developed and what biological implications this has had in characterising the role of IRE1 in GBM pathophysiology and GBM research we shall first go through a body of work (Chapter 4) in the form of a research publication “Dual IRE1 RNase functions dictate glioblastoma development” published in Embo Molecular Medicine (2018 Mar;10(3). pii: e7929) covering a comprehensive study of the potential differential effect IRE1 may have on GBM development through its XBP1 and RIDD signalling. This work not only provides a working model of altered IRE1 activity in patient derived primary cell lines but also lays a significant basis for patient stratification through transcriptomic, clinical, in vitro and in vivo data. In Chapter 5 we will next focus a bit further into the effect of IRE1 in GBM maintenance and the role it may play in the way GBM cells respond to treatment. We have established in the introductory chapters of this thesis that one of the biggest barriers to successful treatment in GBM is tumour heterogeneity and chemotherapy resistance by stem properties. We shall look into why IRE1 is a major player in this phenotypic fluidity and will explore the signalling pathways that mediate this control, thus not only enriching IRE1’s carcinogenic repertoire but also providing clues as to how pertinent IRE1 modulation is in GBM and how important sensitisation to chemotherapy in particular subsets of patients may be. This is in the form of a research manuscript entitled “IRE1 signalling maintains glioblastoma cells differentiation through the XBP1s/miR-148a-mediated repression of stemness transcription factors.” Having established a strong molecular rationale of importance of IRE1 in GBM oncogenesis and maintenance and having provided thus evidence of its validity as a therapeutic target, in Chapter 6 we shall explore a proof of concept novel drug discovery approach that yields new inhibitors of IRE1 activity by targeting its kinase domain. We will see these molecules tested in the GBM models constructed earlier in this thesis and explore their ability to sensitise GBM cells to TMZ. This body of work will be in the form of a research manuscript currently submitted to Molecular Cancer Therapeutics and the novel inhibitors under patent review.

“Results of Targeting IRE1 activity in Glioblastoma Multiforme”. This is a graphical representation of the logical evolution of the results section of this manuscript.

91 Chapter 4

9Dual IRE1 RNase functions dictate glioblastoma development

9 The artwork for the cover of this issue of EMBO Molecular Medicine is owned by EMBOpress.

92 Chapter 4: Foreword

93 Published online: January 8, 2018

Research Article

Dual IRE1 RNase functions dictate glioblastoma development

Stéphanie Lhomond1,†, Tony Avril2,3,†, Nicolas Dejeans1,‡, Konstantinos Voutetakis4,5,‡, Dimitrios Doultsinos2,3, Mari McMahon2,3,6, Raphaël Pineau2,3, Joanna Obacz2,3, Olga Papadodima4, Florence Jouan2,3, Heloise Bourien2,3, Marianthi Logotheti4,7, Gwénaële Jégou2,3, Néstor Pallares-Lupon1, Kathleen Schmit1, Pierre-Jean Le Reste2,8, Amandine Etcheverry9, Jean Mosser9, Kim Barroso2,3, Elodie Vauléon2,3, Marion Maurel2,3,6, Afshin Samali6 , John B Patterson10, Olivier Pluquet11, Claudio Hetz12,13,14,15,16, Véronique Quillien2,3, Aristotelis Chatziioannou4,7 & Eric Chevet2,3,*

Abstract Keywords cancer; endoplasmic reticulum; IRE1; regulated IRE1-dependent decay; XBP1 Proteostasis imbalance is emerging as a major hallmark of cancer, Subject Categories Cancer; Chromatin, Epigenetics, Genomics & Functional driving tumor aggressiveness. Evidence suggests that the endo- Genomics; Neuroscience plasmic reticulum (ER), a major site for protein folding and quality DOI 10.15252/emmm.201707929 | Received 21 April 2017 | Revised 8 December control, plays a critical role in cancer development. This concept is 2017 | Accepted 11 December 2017 | Published online 8 January 2018 valid in glioblastoma multiform (GBM), the most lethal primary EMBO Mol Med (2018) 10:e7929 brain cancer with no effective treatment. We previously demonstrated that the ER stress sensor IRE1a (referred to as IRE1) contributes to GBM progression, through XBP1 mRNA splicing and Introduction regulated IRE1-dependent decay (RIDD) of RNA. Here, we first demonstrated IRE1 signaling significance to human GBM and Glioblastoma multiforme (GBM) is one of the most lethal adult defined specific IRE1-dependent gene expression signatures that cancers, as the majority of patients die within 15 months after diag- were confronted to human GBM transcriptomes. This approach nosis (Anton et al, 2007). GBM is an aggressive, incurable glioma allowed us to demonstrate the antagonistic roles of XBP1 mRNA (grade IV astrocytoma, WHO classification) due to great heterogene- splicing and RIDD on tumor outcomes, mainly through selective ity of cell subtypes within the tumor and to the presence of invasive remodeling of the tumor stroma. This study provides the first spots that cannot be easily cured by surgical resection or targeted demonstration of a dual role of IRE1 downstream signaling in radiation. To limit tumor recurrences from invasive cells, cancer and opens a new therapeutic window to abrogate tumor chemotherapy [temozolomide (TMZ)] was added to surgery and progression. radiation (Stupp et al, 2005). Although this combined therapy has demonstrated some efficiency, it only increases patient’s median

1 Université de Bordeaux, Bordeaux, France 2 INSERM U1242, “Chemistry, Oncogenesis, Stress, Signaling”, Université de Rennes 1, Rennes, France 3 Centre de Lutte Contre le Cancer Eugène Marquis, Rennes, France 4 Institute of Biology, Medicinal Chemistry & Biotechnology, NHRF, Athens, Greece 5 Department of Biochemistry & Biotechnology, University of Thessaly, Larissa, Greece 6 Apoptosis Research Centre, School of Natural Sciences, NUI Galway, Galway, Ireland 7 e-NIOS PC, Kallithea-Athens, Greece 8 Department of Neurosurgery, University Hospital Pontchaillou, Rennes, France 9 Integrated Functional Genomics and Biomarkers Team, UMR6290, CNRS, Université de Rennes 1, Rennes, France 10 Medinnovata Inc., Ventura, CA, USA 11 Institut Pasteur de Lille, CNRS UMR8161 “Mechanisms of Tumourigenesis and Targeted Therapies”, Université de Lille, Lille, France 12 Biomedical Neuroscience Institute, Faculty of Medicine, University of Chile, Santiago, Chile 13 Program of Cellular and Molecular Biology, Institute of Biomedical Sciences, University of Chile, Santiago, Chile 14 Center for Geroscience, Brain Health and Metabolism, Santiago, Chile 15 Buck Institute for Research on Aging, Novato, CA, USA 16 Department of Immunology and Infectious diseases, Harvard School of Public Health, Boston, MA, USA *Corresponding author. Tel: +33 223237258; E-mail: [email protected] † These authors contributed equally to this work as first authors ‡ These authors contributed equally to this work as second authors

ª 2018 The Authors. Published under the terms of the CC BY 4.0 license EMBO Molecular Medicine 10:e7929 | 2018 1 of 19

94 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

survival from 12.1 to 14.6 months. Thus, understanding biological data, obtained using established cell lines, patient tumor samples, processes of GBM progression and treatment resistance represents a and primary GBM lines, depict a complex scenario where IRE1 major challenge to develop more effective therapies. signaling orchestrates distinct aspects of GBM biology, thereby The ER is the major subcellular compartment involved in protein offering novel targets for therapeutic intervention. folding and secretion. Accumulating evidence supports an emerging role of ER proteostasis alterations in cancer development, having been implicated in most hallmarks of cancer (Urra et al, 2016). ER Results stress triggers an adaptive reaction known as the unfolded protein response (UPR), which aims to recover proteostasis or to induce IRE1 activity and human GBM tumor properties apoptosis of irreversibly damaged cells (Walter & Ron, 2011). Several studies in animal models of cancer using genetic or pharma- We previously identified an IRE1-dependent gene expression signa- cological manipulation of the UPR have demonstrated a functional ture in U87 cells using IRE1 dominant-negative-expressing cells, an role of this pathway in cancer (Hetz et al, 2013). The UPR is initi- approach that fully blocks all RNase outputs of this ER stress sensor ated by the activation of three ER transmembrane proteins known (Pluquet et al, 2013). Functional annotation of the genes comprised as PERK, ATF6, and IRE1 (Hetz et al, 2015). IRE1a (referred to as in the IRE1-dependent gene expression signature revealed the IRE1 hereafter) is a serine/threonine kinase and endoribonuclease enrichment in biological functions associated with stress responses, that represents the most conserved UPR signaling branch in evolu- cell adhesion/migration, and with the inflammatory and immune tion, controlling cell fate under ER stress (Hetz et al, 2015). Once response (Fig 1A). This gene expression signature was processed activated, IRE1 oligomerizes thus engaging three major downstream through the Bioinfominer pipeline (Appendix Fig S1) to increase its outputs including the activation of JNK (Urano et al, 2000; Han functional relevance, and this led to the identification of 38 IRE1 et al, 2009), the splicing of XBP1 mRNA (XBP1s) (Yoshida et al, signaling hub genes (Appendix Fig S1). This 38 genes signature was 2001; Calfon et al, 2002), and the degradation of targeted mRNA then confronted to the transcriptomes of the GBM TCGA (Cancer and miRNA, a process referred to as RNA regulated IRE1-dependent Genome Atlas Research Network, 2008) and GBMmark (in-house) decay (RIDD) (Maurel et al, 2014). Importantly, the universe of cohorts (Fig 1B). This analysis revealed the existence of two popula- RIDD targets may depend on the tissue context and the nature of the tions of patients displaying either high or low IRE1 activity, respec- stress stimuli, impacting different biological processes including tively (Fig 1C and Appendix Fig S2). Tumors exhibiting high IRE1 apoptosis, cell migration, and inflammatory responses (Dejeans activity also correlated with shorter survival of the corresponding et al, 2014). Several functional studies have shown that targeting patients (Fig 1D). We then tested the impact of IRE1 signaling on the expression or the RNase activity of IRE1 reduces the progression the expression levels of IBA1, CD14, and CD163 as markers of the of various forms of cancer mostly due to ablating the prosurvival inflammatory/immune response in the tumors (Fig 1E), the levels effects of XBP1 on tumor growth (Chevet et al, 2015; Obacz et al, of CD31 and vWF to monitor angiogenesis (Fig 1F), or RHOA, 2017), and we have previously demonstrated its functional implica- CYR61, and CTGF expression as indicators of tumoral invasion tion in various models of experimental glioblastoma (Drogat et al, (Fig 1G). This revealed that tumors exhibiting high IRE1 activity 2007; Auf et al, 2010; Dejeans et al, 2012; Pluquet et al, 2013; also presented markers of massive infiltration of macrophages, with Jabouille et al, 2015). Moreover, large-scale sequencing studies on high vascularization and invasive properties. Similar observations human cancer tissue samples performed by The Cancer Genome were also obtained when analyzing the GBMmark dataset Atlas (TCGA) initiative (Cancer Genome Atlas Research Network, (Appendix Fig S2B–D). Activation of the IRE1/XBP1 axis was con- 2008; Parsons et al, 2008) revealed the presence of three somatic firmed in those tumors through the analysis of the expression of mutations on the IRE1 gene in GBM leading to the S769F, Q780* XBP1 target genes ERDJ4 and EDEM1 (Appendix Fig S2E). To con- (Greenman et al, 2007), and P336L (Parsons et al, 2008) variants. firm these observations at the protein level in GBM, fresh tumors Although a previous report aimed at understanding the structural presenting high or low IRE1 activity were dissociated and analyzed impact of some of those mutations in IRE1 function (Xue et al, for CD45 and CD11b expression by FACS. This analysis revealed 2011), little is known on how their differential contribution to RIDD that high IRE1 signaling correlated with strong macrophage infiltra- and XBP1 mRNA splicing impacts on GBM development and tion (Fig 1H). Moreover, the presence of endothelial cells in tumors progression. was detected by FACS after CD31 labeling and was increased in Our previous findings indicated that IRE1 also contributes to GBM tumors exhibiting high IRE1 activity (Fig 1I). Finally, tumors mRNA degradation in cancer, having unexpected roles in tissue exhibiting high IRE1 signaling were mainly classified as belonging invasion in GBM, in addition to affecting growth and vascularization to the mesenchymal type of GBM whereas those with low IRE1 (Dejeans et al, 2012; Pluquet et al, 2013). Here, we took advantage activity mostly included pro-neural and classical tumors (Fig 1J). of the selective signaling properties of different IRE1 GBM somatic These data demonstrate that IRE1 activation is found in human mutants and we demonstrate that the modulation of IRE1 signaling tumors and correlate with more aggressive cancers with shorter characteristics in GBM cells controls tumor aggressiveness, not only patient survival. by providing selective advantages to the tumor cells themselves, but also by remodeling the tumor stroma to the benefit of growth. Identification of a novel somatic mutation on IRE1 in Furthermore, we provide evidence supporting a novel concept human GBM where IRE1-downstream signals play antagonistic roles in cancer development, where XBP1s provides pro-tumoral signals, whereas IRE1 activation in tumors could be due to exposure to stressful envi- RIDD of mRNA and miR17 rather elicits anti-tumoral features. Our ronments (nutrient/oxygen deprivation, pH, immune response) but

2 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

95 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

ABC D

EFG

HIJ

Figure 1. IRE1 signaling signatures in glioblastoma multiform. A Functional annotations of the IRE1 gene expression signature identified in U87 cells (Pluquet et al, 2013). B Schematic representation of the analysis workflow. C Hierarchical clustering of GBM patients (TCGA cohort) based on high or low IRE1 activity as assessed with the correlation index of their median z-score with the expression pattern of the IRE1 gene signature of 38 hub genes (see Materials and Methods and Appendix Fig S1). Pearson correlation was used to measure the similarity between different genes and tumor cases, as well. The correlation index refers to the gene expression median z-score with (+) or () sign for identical or reverse expression pattern with that of WT vs. DN, respectively. The expression pattern of WT vs. DN has been described in detail in Pluquet et al (2013). Blue: low correlation index, red: high correlation index. D Survival analysis of the GBM patients exhibiting high (red) or low (green) IRE1 activity. Student’s t-test was used with Welch’s correction when SD different. E–G Expression of microglial/monocyte/macrophage (IBA1,CD14,CD163) (E), angiogenesis (CD31, vWF) (F), and migration/invasion (RHOA, CYR61, CTGF) (G) markers mRNA in the IRE1high (red) and IRE1low (green) populations. Probe analysis was carried out in data from 258 and 265 tumors in IRE1high and IRE1low groups, respectively. Horizontal lines indicate median; box lines indicate first & third quartiles; whiskers indicate min & max. Student’s t-test was used with Welch’s correction when SD different. H FACS analysis of CD45/CD11b in freshly dissociated GBM tumors exhibiting high or low IRE1 activity. I FACS analysis of CD31 in freshly dissociated GBM tumors exhibiting high or low IRE1 activity. In both cases, 7AAD was used to exclude dead cells. J Relative distribution of the different classes of GBM—pro-neural (blue), neural (orange), classic (green), and mesenchymal (red) according to the tumor status, namely IRE1high or IRE1low.

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 3 of 19

96 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

also to the presence of somatic mutations in the IRE1 coding gene. trans-autophosphorylation (Fig 2B and Appendix Fig S4C) as well Previous tumor sequencing studies identified IRE1 mutations that as XBP1 mRNA splicing (Fig 2B). Expression of the Q780* variant were defined as driver in various cancers among which three were also prevented XBP1 mRNA splicing by endogenous IRE1 in found in GBM (Greenman et al, 2007; Parsons et al, 2008). Here, response to tunicamycin treatment (Appendix Fig S4D). In addition, we sequenced the IRE1 gene (ERN1) exons in 23 additional GBM P336L and A414T mutations increased IRE1 oligomerization capac- samples and identified a fourth IRE1 mutation in one GBM human ity (Appendix Fig S4C), leading to IRE1 hyperphosphorylation and sample (Appendix Fig S3A). This somatic A414T mutation came enhanced XBP1 splicing (Fig 2A and B). It is important to note that from an aggressive, mesenchymal-like GBM developed in a 70-year- WT-IRE1 overexpression efficiently increased RIDD activity on old female. Immunohistochemistry staining revealed that this tumor PERIOD1 (PER1), COL6A1, and SCARA3 mRNAs, whereas little was also highly vascularized (CD31 staining) and showed strong impact was observed on other previously reported RIDD substrates XBP1s staining (Appendix Fig S3B). Sequence alignment indicates such as SPARC, PDGFRbeta, and VEGF-A mRNAs, thereby pointing that whereas the mutations P336L, S769F, and Q780* affect toward RIDD selectivity associated with IRE1 variants. As such, the conserved amino acids in various species, the mutation identified in four mutations had different effects depending on the targeted our sequencing study altered an apparently less conserved amino mRNA (Fig 2C). This substrate selectivity might result from modifi- acid, which was only conserved in dog, chimpanzee, and human cations in IRE1 binding to luminal or cytosolic partners due to IRE1 but not in rodents (Appendix Fig S3C). This property could explain overexpression or mutations (i.e., altered oligomerization or signal- why the A414T mutation, previously described in GBM samples, ing properties). Finally, following the observation of Upton and has been excluded from further analyses, as it was considered as a colleagues (Upton et al, 2012), we found that IRE1 variant RNase SNP or a secondary acquired mutation (Cancer Genome Atlas activity controlled miR-17 (miR-17-5p) expression in GBM. Indeed, Research Network, 2008; Parsons et al, 2008). Interestingly since the A414T variant led to increased miR-17 expression under basal the first discovery of IRE1 somatic mutations in cancers in 2007, a conditions while the P336L variant led to low miR-17 levels number of cancer exome or whole-genome sequencing studies have (Fig 2D). IRE1 RNase inhibition mediated by MKC4485 (Volkmann also reported around 50 mutations but none of them in GBM et al, 2011) restored the expression of miR-17 in P336L IRE1 variant (Chevet et al, 2015). expressing U87 cells (Appendix Fig S4E), thus confirming the involvement of IRE1 RNase in miR-17 expression. Tunicamycin- Different kinase and RNase activities of IRE1-related induced ER stress engaged IRE1 activation and led to further miR-17 cancer variants degradation (Appendix Fig S4F). We have summarized the differential impact of IRE1 variants on distinct downstream signaling IRE1 is a bifunctional protein that contains a kinase and a RNase outputs in Fig 2E. domain involved in three downstream signaling pathways including (i) activation of stress pathways [i.e., JNK and NFKB (Hetz, 2012)], U87 phenotype and signaling upon expression of IRE1 variants (ii) the degradation of targeted RNAs (RIDD), and (iii) the unconventional splicing of XBP1 mRNA. The localization of IRE1 muta- To further investigate the impact of IRE1 variant expression in U87 tions found in cancer revealed no apparent clustering of the cells, we first evaluated the cellular phenotypes generated using mutations in the secondary structure, not even into IRE1 catalytic phase microscope imaging. All the cells presented a mesenchymal domains. However, the “cytosolic” mutations S769F and Q780* are phenotype comparable to U87 transfected with an empty vector located in the kinase domain of the protein whereas the “luminal” (Fig 2F) or parental U87 (not shown), except for those expressing mutations P336L and A414T are located in putative alpha-helical the IRE1 P336L variant, which displayed an epithelial-like pheno- domains (Appendix Fig S3D). To measure the potential impact of type. These cells exhibited similar proliferation rates (Appendix Fig the four mutations found in GBM, we overexpressed either the wild- S5A) and still had the capacity of forming spheres in culture type (WT) or the mutated forms of IRE1 in U87 cells, in a normal (Appendix Fig S5B) as described previously (Dejeans et al, 2012). endogenous IRE1 background (Appendix Fig S4A). The four vari- To further evaluate the IRE1-dependent signaling aspects in GBM, ants were overexpressed in U87 cells using a lentivirus system, and we used the KEGG pathway for glioma that compiles the main as anticipated, the stop mutation Q780* leads to overexpression of a actors involved in gliomagenesis, and identified the components shorter IRE1 protein (80 kDa instead of 110 kDa). Finally, that were previously shown to be directly or indirectly regulated by immunofluorescence studies showed that IRE1 staining co-localized IRE1. This revealed that 45% of the components comprised in this with an ER marker (KDEL staining) and thereby confirmed that pathway were controlled through IRE1-dependent mechanisms mutations did not affect IRE1 localization to the ER (Appendix Fig (Appendix Fig S5C; yellow boxes). We then monitored the expres- S4A and B). sion levels of PDGFRbeta and p53 (respectively top and bottom of As reported in other cellular system (Han et al, 2009), the the pathway, Appendix Fig S5D). PDGFRbeta expression was highly overexpression of the WT form in U87 was sufficient to activate expressed in cells expressing EV and the P336L mutant. Interest- IRE1 in basal conditions compared to the control empty vector ingly, the expression of wild-type p53 in U87 cells (Cerrato et al, (EV)-expressing cells, as indicated by basal IRE1 phosphorylation, 2001) was upregulated by 15-fold in P336L-expressing cells. In those as well as XBP1 mRNA splicing (Fig 2A and B). As expected, Q780* cells, p53 mRNA was not altered compared to EV cells corresponded to a loss-of-function mutation regarding the splic- (Appendix Fig S5E) and no mutations were found by sequencing ing of XBP1 mRNA. Indeed, the loss of the last fragment of the (not shown), thereby suggesting a translational regulation of the kinase domain and the entire RNase and C-terminus domains did protein. To further characterize the impact of the IRE1 variants on not affect IRE1 oligomerization but impaired the resulting cell signaling pathways, we used a transcriptomic approach.

4 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

97 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

D G

Figure 2.

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 5 of 19

98 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

◀ Figure 2. Impact of somatic mutations on IRE1 signaling. A Anti-IRE1 Phostag immunoblot showing both phosphorylated (p-IRE1) and non-phosphorylated (IRE1) IRE1 proteins revealed IRE1 phosphorylation in basal conditions due to overexpression of WT, P336L and A414T but not S769F nor Q780* forms of IRE1. EtBr-stained agarose gel of XBP1 cDNA amplicons corresponding to unspliced (XBP1u) and spliced (XBP1s) forms of XBP1 mRNA revealed XBP1 splicing in basal conditions due to overexpression of WT, S769F, P336L and A414T but not Q780* forms of IRE1. B Bar graph representing the quantification of three levels of IRE1/XBP1 activation: IRE1 phosphorylation (p-IRE1/IRE1) and XBP1 mRNA splicing (XBP1s/ (XBP1u + XBP1s)) measured as indicated in (A). Three independent biological samples were used. Data from five independent experiments are presented as means SD. P-values are indicated. Student’s t-test was used. C Heat-map representation of RIDD of mRNA target expression (normalized to 18S) after 2-h Actinomycin D (ActD) treatment. Three independent biological samples were used. D Analysis of miR-17-5p expression by RT-qPCR in IRE1 variant expressing cells under basal conditions. Data from three independent experiments are presented as means SD. Student’s t-test was used. E Recapitulative table of the signaling properties of IRE1 WT, DN, P336L, and A414T variants. F Phenotypic characterization of U87 cells expressing the different IRE1 variants and imaged by phase contrast microscopy. Scale bar = 100 lm. G Heat-map representation of the transcriptomes of U87-expressing variants upon basal conditions or ER stress (induction by tunicamycin). H Schematic representation of the signaling pathway enrichment based on the transcriptome data.

Hierarchical clustering revealed that WT IRE1 grossly behaved as variants, which exhibited similar gain of function on IRE1 in vitro, the S769F, Q780*, and P336L variants under basal and stress condi- showed diametrically opposed behaviors in vivo on tumor develop- tions. In contrast, cells expressing the IRE1 A414T variant exhibited ment. Indeed, whereas P336L totally blocked tumor formation, a very different gene expression profile than the other cell types that A414T shortened mouse survival (Fig 3B), most likely by promoting more closely resembled the signature observed in IRE1-DN cells tumor growth and vascularization with hallmarks of vessel cooption (Pluquet et al, 2013). The expression profiles were then analyzed (Fig 3A, bottom). Interestingly, tumors formed from EV, WT, and for signaling pathway activation and unveiled possible pathways A414T cells showed high XBP1s expression as assessed by immuno- selectively activated by IRE1-related cancer variants (Fig 2G). Func- histochemistry which did not account for the differences observed tional analysis of the gene enrichment pattern indicated a major in mouse survival (Fig 3C). Remarkably, the pro-angiogenic effects impact of IRE1 A41AT mutation on signaling pathways involved in of the A414T mutation not only increased the number of vessels metabolism control, extracellular matrix (ECM) organization, and associated with the tumor mass but also increased the size of those cell homeostasis maintenance, whereas IRE1-DN impacted mostly vessels (Fig 3D and E), an effect that was much less visible in early genes related to ECM organization, cell homeostasis, and the steps of tumorigenesis (Fig 3A and E). Furthermore, the impact of immune system. Interestingly, the impact of other variants on basic the A414T mutation on the immune infiltrate to the tumor site was cellular signaling functions remained limited compared to the also evaluated in vivo and showed that expression of this IRE1 vari- pattern elicited by IRE1 A414T and DN expression (Fig 2H and ant in U87 cells resulted in the formation of tumors presenting very Appendix Table S1). low levels of macrophage infiltration (F4/80 staining; Fig 3F). This was not the case for other variants (Fig 3G). Finally, tumor- Modulation of tumor development in vivo upon expression of infiltrating spots were quantified as previously described (Pluquet IRE1 variants et al, 2013) and showed major infiltration/invasion of DN as previously observed (Drogat et al, 2007; Auf et al, 2010; Jabouille et al, To evaluate the significance of each IRE1 variant to tumor growth 2015) but also a significant positive impact of the expression of the in vivo, we implanted control U87 or cells expressing WT and A414T variant (Fig 3A and H). Thus, our results suggest that selec- mutated forms of IRE1 into mouse brain, as previously described tive genetic alterations affecting IRE1 activity condition the specific (Auf et al, 2010; Pluquet et al, 2013). Fifteen days post-implanta- biological outputs observed at the level of tumor growth, survival, tion, five animals of each group were sacrificed and brains isolated angiogenesis, and immune cell infiltration. for immunofluorescence (IF) staining of tumor cells (vimentin) and vessels (CD31). As expected, IRE1 overexpression impacted tumor IRE1 downstream signals drive changes in the growth and vascularization, whereas impairment of IRE1 signaling tumor microenvironment (IRE1-DN) reduced both size and vascularization of the tumors (Fig 3A). An exception of this tumorigenic effect of IRE1 was To further dissect the contribution of signals downstream of IRE1 in observed with the P336L mutation. Indeed, injection of U87 express- GBM tumor phenotypes, we took advantage of the properties of ing the IRE1 P336L never led to the formation of a visible tumor IRE1 variants. Global analysis of our results highlights the signaling (> 15 injections). This phenomenon may be a consequence of the differences driven by IRE1 mutant forms where WT IRE1, the observed overexpression of the tumor suppressor p53 in those cells P336L, and A414T variants exhibited high XBP1 mRNA splicing and (Appendix Fig S5), leading to the attenuation of U87 aggressive RIDD activities whereas the expression of miR-17 was elevated in phenotype. cells expressing WT or IRE1-DN as well as those expressing the Among the four mutations, the loss-of-function mutations S769F A414T variant (Figs 2E and 4A). Based on this analysis, we and Q780* appeared to have little effects on mouse survival reasoned that genes whose expression was upregulated in WT, (Fig 3B); however, the Q780* mutation accelerated the early steps P336L, and A414T could reflect targets under the control of XBP1s. of tumor growth compared to the control tumors (orange vs. black Using the transcriptome profiles established in Fig 4, we identified a lines; Fig 3B). Remarkably, expression of the P336L and A414T list of 40 genes that segregate with high XBP1s levels (Fig 4B and

6 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

99 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

D F

EGH

Figure 3.

Appendix Table S2). We then used the XBP1s signature to classify or low expression of XBP1s target genes (Fig 4C). This revealed that tumors from the GBMmark cohort (Appendix Fig S6A) and analyzed the three markers studied showed higher expression levels in the expression of IBA1, CD31, and RHOA in tumors exhibiting high tumors with high XBP1s target gene expression. This was further

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 7 of 19

100 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

◀ Figure 3. Impact of IRE1 somatic mutations on tumor development. A Tumor cells (U87) were injected into the brain of recipient mice (5 to 15 Rag-c 2C/ per condition). Animals were sacrificed 15 days post-injection. Brains were collected and analyzed by immunofluorescence with anti-vimentin (green) and anti-CD31 (red) antibodies (scale bar = 500 lm). B Tumor cells (U87) were injected into the brain of recipient mice (Rag-c 2C/). Animals were sacrificed at first clinical sights of tumor development, and each sacrifice was reported in the Kaplan–Meier curve, indicating a gain of lethality for tumors formed in WT or A414T conditions. C Brains were collected and analyzed by H&E and immunostaining with anti-XBP1s antibodies (scale bar in top row = 50 lm; scale bar in middle row = 150 lm; scale bar in bottom row = 700 lm). T: tumoral tissue; NT: non-tumoral tissue. DCD31 staining (red) and nuclear nucleus staining (blue) in tumors collected at sacrifice exemplified with control (EV) and A414T-expressing cells (scale bar = 250 lm). E Graphic representation of tumor vascularization for each tumor. Tumors were classified into four groups relative to their degree of vascularization: (i) avascular tumors (no): no apparent blood vessels in the tumor; (ii) poorly vascularized (low): A few blood vessels are seen in the tumor; (iii) moderately vascularized (medium): Numerous blood vessels are seen in the tumor; (iv) highly vascularized (high): The tumor surface is covered with blood vessels (as well as wide). For animal experimentation, data shown are the mean SEM of five tumors per experiment. FF4/80 (macrophage) staining in tumors collected at sacrifice exemplified with control (EV) and A414T expressing cells. Scale bar = 100 lm. G Quantification of macrophage infiltration in tumors, data are represented as the mean SEM of five tumors per experiment. Statistics were determined using two-way ANOVA with Bonferroni post-test. H Quantification of tumor invasion as described previously. Tumors were stained with anti-vimentin antibodies, and the number of vimentin-decorated spots was quantified per surface area as previously described (Pluquet et al, 2013). Data are the mean SEM of five tumors per experiment. Statistics were determined using two-way ANOVA with Bonferroni post-test.

confirmed using immunohistochemistry with antibodies against evaluate tumors with high or low IRE1/miR-17 in the GBMmark XBP1s and IBA1. A total of 35 cases of GBM were analyzed with cohort and to monitor the expression of IBA1, CD14, CD31, vWF, anti-XBP1s and revealed either no staining (Fig 4D–1) or staining in RHOA, and CTGF (Fig 5C). As for RIDD of mRNA, these data indi- the nucleus (Fig 4D–2) and in the cytoplasm (Fig 4D–3/4). A subset cated that RIDD for miR-17 exhibited anti-angiogenic and anti- of those tumors (n = 24) was then analyzed for IBA1 expression migratory effects. This led us to correlate high RIDD IRE1 activity, (Appendix Fig S6B), and a correlation between the presence of IBA1 which might lead to low miR-17 expression, and better outcome in and that of XBP1s was established thereby indicating that high GBM patients. To test this hypothesis, we evaluated the expression XBP1s in the tumor may control immune cell (macrophage) infiltra- of miR-17-5p in 30 GBM tumors and identified two groups of tumors tion (Fig 4E). exhibiting low or high miR-17-5p expression (Fig 5D). Patient Next, we investigated how RIDD activity could impact on tumor survival was evaluated in those two groups of patients and revealed characteristics. To this end, we determined a potential RIDD signa- that low miR-17-5p levels in tumors correlate with better survival ture based on the ability of IRE1 to cleave select mRNA in vitro. This than those patients presenting high miR-17-5p tumors (Fig 5E), screening identified a group of 1,141 mRNAs susceptible to be thereby confirming our initial hypothesis. To functionally explore cleaved in vitro by IRE1 (Appendix Fig S7A and Table S3), which the role of the IRE1/miR-17 axis in GBM development, we blunted were then intersected with the set of genes upregulated in IRE1-DN miR-17 activity with anti-miR-17 sponges in EV and A414T cells and cells. This analysis yielded a subset of 37 potential GBM-specific tested the expression of predicted miR-17 target genes. We con- RIDD targets (Fig 4F and Appendix Table S4). Their functional firmed that antagonizing miR-17 increases the expression of miR-17 annotation suggests the enrichment in genes involved in the NOD targets in cells expressing the A414T IRE1 variant (Fig 5F). To pathway, interaction with the environment, and biogenesis of cellu- further evaluate the functional role of miR-17 in GBM, we moni- lar components (Appendix Fig S7B). Then, this cluster of mRNAs tored the impact of either XBP1 silencing or antagonizing miR-17 in was used to identify RIDD-positive and RIDD-negative tumor popu- U87 or U251 cells expressing an empty vector or the IRE1 A414T lations in the GBMmark cohort (Fig 4G). The expression of immune variant. This approach revealed that siRNA-mediated XBP1 silencing infiltration, angiogenesis, and invasion markers in these populations in U87 cells impaired monocyte chemoattraction (Fig 5G) whereas confirmed previous results and ruled out tumoral RIDD of mRNA in antagonizing miR-17 did not have any significant effect (not the recruitment of immune cells (Fig 4H). In summary, in contrast shown). In addition, we found that overexpression of IRE1 A414T in to the IRE1/XBP1 axis that exhibits pro-tumorigenic signaling U251 cells increased cell migration using a trans-well assay, a features, the RIDD of mRNA pathway may antagonize tumor inva- feature that was impaired by XBP1 silencing or miR-17 buffering sion and angiogenesis with no significant effect on immune cells (Fig 5H). infiltration. Our data led us to propose a complex model in which the IRE1/ XBP1 signaling axis would promote GBM aggressiveness through Differential contribution of RIDD and XBP1 mRNA splicing the enhancement of tumor immune infiltration and angiogenesis as to GBM well as tumor cell invasiveness properties, whereas RIDD (of mRNA and miRNA) would play an anti-tumoral role by selectively reducing We then investigated the role of IRE1/miR-17 axis in cancer progres- tumor angiogenesis as well as tumor cell invasiveness (Fig 6A). To sion. To this end, we took advantage of the properties of the dif- further demonstrate the divergent activities of IRE1 in cancer, the ferent IRE1 variants toward miR-17 (Fig 4A) and established a TCGA cohorts (microarrays and RNAseq) were analyzed for popula- minimal group of genes whose expression could be under the tions exhibiting low and high XBP1 splicing and RIDD activities. control of miR-17 (Fig 5A and Appendix Table S5). This set of genes Hierarchical clustering revealed four major groups as follows is involved in morphogenetic programs, cell adhesion, synthesis of XBP1shigh/RIDDlow (XBP1+/RIDD ); XBP1slow/RIDDlow (XBP1 / aromatic compounds, and to a lesser extent in the response to reac- RIDD ); XBP1slow/RIDDhigh (XBP1 /RIDD+); and XBP1shigh/ tive oxygen species (Fig 5B). This information was then used to RIDDhigh (XBP1+/RIDD+) (Fig 6B and C, see Materials and

8 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

101 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

ABC

DEF

Figure 4. XBP1s and RIDD of mRNA signals in GBM. A Schematic representation of the signaling characteristics of IRE1 variants based on data generated in Fig 2, toward XBP1 mRNA splicing, RIDD of mRNA, and RIDD of miR-17. B Intersection of the upregulated genes in WT, P336L, and A414T-expressing cells as the hallmark of XBP1s expression. A list of 40 genes was established and confronted to the tumor transcriptomes of the GBMmark cohort (Appendix Table S2). XBP1s high and XBP1s low groups of tumors were established. C Expression analysis of IBA1,CD31, and RHOA mRNA in the XBP1s high (red; n = 44) and low (green; n = 75) groups of tumors. Horizontal lines indicate median; box lines indicate first & third quartiles; whiskers indicate min & max. Student’s t-test was used with Welch’s correction when SD different. D Characterization of XBP1s high and low tumors using immunohistochemistry (no XBP1s expression (1), and increasing amounts of XBP1s expression (2–4) are shown). E Correlation of XBP1s and IBA1 staining. F Intersection of the list of RNA cleaved by IRE1 in vitro (Appendix Table S3) and that of mRNA whose expression is upregulated in IRE1-DN cells under basal conditions. G A list of 37 genes was established and confronted to the tumor transcriptomes of the GBMmark cohort. RIDDhigh and RIDDlow groups of tumors were established. H Expression analysis of IBA1,CD31, and RHOA mRNA was evaluated in the RIDDhigh (red; n = 64) and RIDDlow (green; n = 55) groups of tumors (I). Horizontal lines indicate median; box lines indicate first & third quartiles; whiskers indicate min & max. Student’s t-test was used with Welch’s correction when SD different.

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 9 of 19

102 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

ADB E

C F

Figure 5. RIDD of miRNA (miRIDD) signals in GBM. A Intersection of the downregulated genes in WT, DN, and A414T-expressing cells (high miR-17) yielding a list of 38 genes. B These 38 genes were analyzed for being predicted direct miR-17 targets, which reduced the list to 12 genes (Appendix Table S5). Their functional annotation is shown. C The 12 genes list was confronted to the tumor transcriptomes of the GBMmark cohort. miRIDDhigh and miRIDDlow groups of tumors were established, and expression of IBA1,CD14,CD31, vWF, and RHOA, CTGF mRNA was evaluated in the miRIDDhigh (green; n = 10) and miRIDDlow (red; n = 18) groups of tumors. Horizontal lines indicate median; box lines indicate first & third quartiles; whiskers indicate min & max. Student’s t-test was used. D Expression of miR-17-5pin30 GBM tumors and distribution of these tumors in low (red) and high (green) miR-17-5p groups. Student’s t-test was used. E Kaplan–Meier survival curves of low (red) and high (green) miR-17-5p GBM tumor patients. Student’s t-test was used. F Real-time PCR analysis of three miR-17 targets identified in our study (TSLP, LMO2, MAN1C1) and a positive control (PTEN) on mRNA extracted from control or IRE1 A414T expressing U87 cells. Student’s t-test was used. G Monocyte chemoattraction with medium conditioned by U87 cells expressing an empty vector or the IRE1 A414T variant and silenced or not for XBP1. Student’s t-test was used. *P-value = 0.0015,**P-value = 0.0009. H Glioblastoma cell (U251) migration assay in Boyden chamber carried out with cells expressing or not IRE1 A414T and either silenced for XBP1 or transfected with an antago-miR-17. Student’s t-test was used. Data information: Data from five independent experiments are presented as means SD.

Methods). Remarkably, patient survival analysis in the XBP1 / This was also confirmed by the fact that the XBP1+/RIDD popula- RIDD+ and XBP1+/RIDD populations revealed a clear segregation tion was enriched in mesenchymal tumors whereas XBP1 /RIDD+ where the former group survived statistically longer than the latter comprised more pro-neural and neural tumors (Appendix Fig S7C). as suggested from our working model (Fig 6D and E). Moreover, as Interestingly, when the splicing of XBP1 was evaluated in the TCGA anticipated, XBP1+/RIDD GBM exhibited higher expression IBA1, RNAseq cohort and compared to the XBP1+ groups established CD31, and RHOA than the XBP1 /RIDD+ tumors (Fig 6F and G). using our method, we correlated both information thereby

10 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

103 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

confirming the validity of our approach (Appendix Fig S7D). Finally, related to cell migration (EMT) and chemokine production. This differences in the survival of the patients belonging to the intermedi- revealed that overexpression of WT IRE1 increased the expression ate groups (i.e., XBP1 /RIDD ; XBP1+/RIDD+) were not statisti- of the EMT-related genes VIM, ZEB1, and TGFB2 whereas that of cally significant (Appendix Fig S7E). Taken together, these Q780* IRE1 did not affect it when compared to the control parental experiments suggest an antagonistic role of XBP1 mRNA splicing line (Fig 8D). Similar results were also observed for the expression and RIDD in GBM specifically regarding cell migration and angio- of the chemokines CXCL2, CCL2, and IL6 (Fig 8E), thereby con- genesis. firming the contribution of IRE1 signaling to those pathways. We then monitored the functional effects of IRE1 modulation in Modeling IRE1 contribution to GBM development in primary primary lines on tumor cell migration since the cellular phenotypes GBM lines were altered when expressing IRE1 WT or mutant (Appendix Fig S8B). This was carried out using Boyden chamber-based assays We then tested whether the tumor classification established in Fig 6 (Fig 8F), and quantitation showed that in most cases IRE1 activation was also relevant in primary GBM lines that could in turn serve as promoted cell migration whereas IRE1 inhibition completely abro- an in vitro model for better understanding the role of IRE1 signaling gated it in all primary lines tested (Fig 8G). As we previously in GBM development. We therefore applied the same clustering showed that IRE1/XBP1 signaling was also involved in monocytes method as in Fig 6 on the transcriptome datasets from 12 primary chemoattraction, we tested whether modulating IRE1 activity in GBM lines. This revealed that the 12 lines clustered into the same primary lines would affect this property using Boyden chamber- four groups as observed for the tumors (Fig 7A). These tumor lines based migration and FACS analyses (Fig 8H). Quantitation of the exhibited various phenotypes in culture, notably regarding adhe- results showed that, as observed for migration, enhancing IRE1 sion/protrusion/migration (Fig 7B). Interestingly, high adhesion/ activity led to increased chemoattraction whereas blunting XBP1s migration also correlated with the XBP1+ status. This was con- through the expression of IRE1 Q780* resulted in decreased capacity firmed using RT-qPCR with lines belonging to the XBP1+ group to attract monocytes in the four primary lines tested (Fig 8I). These exhibiting the highest XBP1s levels (Fig 7C). This information results obtained in primary GBM lines confirm our previous obser- however was not further correlated with the in vitro migration prop- vations in U87 cells and in human tumor samples and support our erties of the primary lines (Fig 7D). To further evaluate the rele- model in which IRE1 activity could control the specific properties of vance of the classification, the 12 lines were orthotopically injected GBM tumor cells through the combined action of XBP1s and RIDD. in mice and the resulting tumors evaluated using vimentin staining (Fig 7E). Together with the tumor size (Fig 7F) and the survival data (Fig 7G), these experiments confirmed that XBP1+/RIDD Discussion tumor cells yielded the most aggressive tumors whereas injecting XBP1 /RIDD+ tumor cells resulted in very small tumors, thereby Our work demonstrates that in GBM IRE1, downstream signals, confirming the results observed in patients’ tumors (Fig 6). We then including XBP1 mRNA splicing and RIDD, dictate tumor phenotypes monitored both macrophage recruitment to the tumors as well as and patient outcomes. At first, we showed the relevance of IRE1 angiogenesis and showed that although high XBP1s correlated with signaling in GBM from two independent cohorts (TCGA and important macrophage infiltration in the tumors (Fig 7H), tumoral GBMmark) and found that high IRE1 activity correlates with shorter large vessel content did not significantly change in the different patient survival and increased tumor infiltration by immune cells, groups (Fig 7I). Interestingly, the expression levels of CCL2 corre- increased tumor angiogenesis, and enhanced invasion/migration lated with high XBP1s (Fig 7J) whereas that of VEGF did not properties of the tumor cells (Fig 1). Previous studies demonstrated (Fig 7K). These data show that primary GBM lines recapitulate, at the importance of IRE1 signaling for tumor aggressiveness; least partially, the IRE1 signaling properties observed in human however, they did not provide any information on the underlying tumors and maintain the expected biological outputs even in vivo. molecular mechanisms involved in this phenomenon. Previous We then further tested whether altering IRE1 activity in those studies established the IRE1 molecular signature using various cells could impact on their tumoral properties. To this end, we over- exogenous acute stresses, such as hypoxia and nutrient deprivation expressed IRE1 WT and the Q780* mutant [known to impair XBP1s (Pluquet et al, 2013). However, the use of such micro-environ- (Fig 2B)] in four primary lines, namely RNS85, RNS87, RNS96, and mental challenges did not recapitulate the complexity of brain RNS130. The lines were selected on the basis of IRE1 mRNA expres- cancer as an experimental mean to define the molecular mecha- sion (Appendix Fig S8A) and then analyzed by Western blot with nisms downstream of IRE1 involved in tumor growth. Thus, we anti IRE1 antibodies (Fig 8A). As expected, IRE1 expression was reasoned that mutations identified on IRE1 in GBM could serve as higher upon overexpression of IRE1 WT and the shorter form of interesting tools to characterize how specific IRE1-dependent signal- IRE1 observed upon expression of the Q780* variant. These dif- ing pathways could control tumor phenotypes at both the tumor cell ferent lines were then monitored for the splicing of XBP1 mRNA and stroma levels. IRE1 mutations in GBM were previously reported (Fig 8B) and for the expression of UPR target genes (Fig 8C). Again, (Chevet et al, 2015), but the functional consequences of those muta- these analyses confirmed that in RNS85, 87, 96, and 130, overex- tions on IRE1 signaling remained undocumented. To further expand pression of IRE1 increased IRE1 activity, thereby resulting in an the repertoire of IRE1 mutations in GBM, we sequenced the IRE1 increased expression of select target genes whereas the expression gene in 23 additional GBM tumors (Appendix Fig S3), and then of the Q780* variant resulted in blunting XBP1s signaling. To further analyzed the signaling characteristics of the variants identified. investigate the role of IRE1 modulation in those lines, we first tested In the present study, we have characterized in detail IRE1 how IRE1 activation or inhibition affected the expression of genes somatic variants using complementary approaches and uncovered a

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 11 of 19

104 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

ABC

E G

Figure 6. Deconvolution of IRE1 signaling in GBM tumors. A Schematic representation of the antagonistic XBP1s and RIDD signals in GBM tumors and their biological impact on tumor aggressiveness. B, C Hierarchical clustering of GBM patients (TCGA cohort—microarray, B; TCGA cohort—RNAseq, C) based on XBP1s and RIDD scores (see Materials and Methods). D, E Kaplan–Meier survival curves of XBP1+/RIDD (red) and XBP1/RIDD+ (blue) GBM patients of the TCGA microarray cohort (D), TCGA RNAseq cohort (E). Student’s t- test was used. F, G Expression of monocyte (IBA1), angiogenesis (CD31), and migration/invasion (RHOA) markers mRNA in the four groups established in the two cohorts. Horizontal lines indicate median; box lines indicate first & third quartiles; whiskers indicate min & max. Student’s t-test was used.

novel mutation in IRE1 on the A414 residue. Expression of this vari- orthotopic tumors (Figs 3 and 7). This result was difficult to explain ant in U87 cells led to highly aggressive cancer with enhanced only on the basis of the signaling characteristics of this mutant vascularization and reduced infiltration of macrophages to the (Fig 2) and suggested a highly complex integrated IRE1 signal

12 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

105 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

FH J

GI K

Figure 7.

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 13 of 19

106 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

◀ Figure 7. Primary GBM lines exhibit IRE1 signaling properties of the parental tumors. A Hierarchical clustering of 12 GBM primary lines based on XBP1s and RIDD scores (see Materials and Methods). B Phase contrast images of the phenotypes exhibited by the primary lines from the four groups established in (A) (scale bar = 100 lm). C Quantitation of XBP1s mRNA in all primary lines relative to the group to which they belong (n = 3, mean SD). ANOVA was used for statistical analyses. D Quantitation of the Boyden chamber migration assays for all the lines (n = 3, mean SD). ANOVA was used for statistical analyses. E GBM primary cell lines were implanted in nude mice. Animals were sacrificed when first clinical signs appeared. Brains were then collected and analyzed for vimentin expression by immunohistochemistry (scale bar = 800 lm). F Quantitation of tumor size. ANOVA was used for statistical analyses. For each cell line tested (represented by a single point on the graph), the average of three independent experiments is shown SD. G Quantitation of mouse survival. ANOVA was used for statistical analyses. For each cell line tested (represented by a single point on the graph), the average of three independent experiments is shown SD. H Quantitation of macrophage infiltration in orthotopic tumors (IBA1 staining). ANOVA was used for statistical analyses. For each cell line tested (represented by a single point on the graph), the average of three independent experiments is shown SD. I Quantitation of angiogenesis in orthotopic tumors (CD31 staining). ANOVA was used for statistical analyses. For each cell line tested (represented by a single point on the graph), the average of three independent experiments is shown SD. J, K Expression of CCL2 (J) and VEGF-A (K) mRNA as determined using RT-qPCR in the different primary GBM lines. ANOVA was used for statistical analyses. For each cell line tested (represented by a single point on the graph), the average of three independent experiments is shown SD.

accounting for both XBP1 mRNA splicing and RIDD characteristics of each distinct IRE1 signaling output. This analysis revealed that to produce the observed tumor phenotype. This was reinforced by the XBP1s signaling axis promoted tumor infiltration by immune our work on the P336L mutation which is, so far, the only IRE1 cells, increased angiogenesis, and enhanced the expression of migra- mutation identified in more than one tumor sample and even in tion/invasion markers. Interestingly, these properties were also con- more than one cancer type [one in glioma (Parsons et al, 2008) and firmed using approaches relying on immunohistochemical analyses two in intestinal cancers (Cosmic)], thereby confirming its relation- on a different subset of tumors (Fig 4). In contrast, RIDD activity ship with cancer development. We thus hypothesized that the onco- (toward either mRNA or miR-17) attenuated both the angiogenic genic potential of this mutation may need a particular cancer response and the migration/invasion properties of the tumor cells context, such as the presence of acquired mutation in key genes for (Figs 4 and 5). The opposite signals elicited by XBP1s and RIDD GBM development (EGFR, PTEN, TP53, NF1, and IDH1), as no confer specific aggressive features to tumors with XBP1+/RIDD previous study defined P336L as a driver mutation. We showed that associated with a worse prognostic outcome than those with TP53 was overexpressed in IRE1-P336L expressing U87 cells which XBP1 /RIDD+ properties (Fig 7). Interestingly, tumors with low most likely promoted tumor suppression in vivo (TP53 wild type in RIDD/XBP1s and those with high XBP1s/RIDD did not yield dif- U87; Appendix Fig S5). This observation rules out the driver role of ferent prognoses in terms of patient survival (Appendix Fig S6) most this mutation that would only subsist in a P53 mutant background. likely due either to the major contribution of other pathways still to Moreover, recent work reported a direct role of the IRE1 target JNK be identified (i.e., EGFR, P53) or to compensatory mechanisms of in stabilizing EGFR ligand epiregulin (EREG) and consequently an both pathways, respectively. Beyond the characterization of these autocrine activation loop of EGFR, which should provide prolifera- signaling properties in U87 cells and in human tumors, we also tive advantage of GBM cells in which EGFR signaling was already demonstrated that those pathways were also conserved in primary altered by mutations (Auf et al, 2013). This hypothesis could also GBM lines and that their modulation had a significant impact on the explain the proliferative effects of A414T mutation, and future stud- tumor cells phenotypes (Figs 7 and 8). These pathology-relevant ies should define the involvement of EGFR or other key GBM tools will now be useful to define how IRE1 signals toward XBP1s proteins in IRE1-dependent GBM growth, as both P336L and A414T or RIDD are regulated at the level of IRE1 and how those two signal- mutations seemed to stabilize IRE1 kinase and RNases activities. ing arms quantitatively interact to drive specific tumor characteris- Hence, the signaling characteristics of IRE1 variants confer GBM tics. Moreover, RIDD of miRNA (miRIDD) activity could interact tumors with specificities that could lead to aggressive features. with known pro-tumoral pathways in GBM through regulation of We took advantage of the signaling characteristics of IRE1 vari- other IRE1 miRNA targets such as miR-34a (Genovese et al, 2012). ants associated with GBM and the fact that we characterized the In addition to providing the first evidence of the co-existence of transcriptome of U87 cells expressing those variants under basal antagonistic IRE1 downstream signals in GBM and correlating those and ER stress conditions to delineate specific molecular signatures with tumor aggressiveness features, our work highlights the

Figure 8. Modulating IRE1 activity in primary GBM lines and phenotypic outcomes. ▸ A Western blot analysis of the expression of WT or Q780* IRE1 in RNS85, RNS87, RNS96, and RNS130 primary GBM lines. B Analysis of XBP1 mRNA splicing using RT-PCR. C Quantitation of the expression of CHOP, HERPUD1, ERDJ4, GADD34 (UPR genes) using RT-qPCR in the four lines. D Quantitation of the expression of vimentin (VIM), ZEB1, TGFB2 (EMT genes) using RT-qPCR in these IRE1-modified lines. E Quantitation of the expression of CXCL2, CCL2, IL-6 (chemokine genes) using RT-qPCR in these modified lines. F Representative phase contrast images of migrating RNS85-derived lines tested in the Boyden chamber migration assay (scale bar = 10 lm). G Quantitation of tumor migration for all the lines (n = 3, mean SD). Student’s t-test was used for statistical analyses. H Representative flow cytometry dot plot graphs obtained in the monocytes chemoattraction assay using RNS85-derived lines conditioned media. I Quantitation of the monocytes chemoattraction for all the lines (n = 3, mean SD). Student’s t-test was used for statistical analyses. Data information: Numbers shown on the top of the graphs presented in (C, D, E) indicate P-values, Student’s t-test was used.

14 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

107 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

AB C

F G

Figure 8.

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 15 of 19

108 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

possibility of using IRE1-targeted therapeutics in cancer. Indeed, it (Miltenyi Biotec, Paris, France). RNS cells (neurospheres enriched is likely that in XBP1+/RIDD tumors, inhibition of IRE1 RNase in cancer stem cells) were directly cultured in DMEM/Ham’s F12 with small molecules (Hetz et al, 2013) or selective inhibition of the (Lonza, Verviers, Belgium) supplemented with B27 and N2 addi- XBP1 mRNA ligase RtcB (Jurkin et al, 2014; Kosmaczewski et al, tives (Invitrogen, Cergy Pontoise, France), EGF (20 ng/ml), and 2014; Lu et al, 2014; Ray et al, 2014) might lead to significant FGF2 (20 ng/ml) (Peprotech, Tebu-Bio). impairment of tumor growth. This type of strategy would be highly relevant for tumor cells exhibiting constitutive expression of XBP1s IRE1 sequencing such as triple-negative breast cancers (Chen et al, 2014). Similarly, an approach aiming at increasing RIDD activity in XBP1 /RIDD+ IRE1 exon sequencing was performed by Beckman Coulter Genomics tumors to induce cell death mechanisms would be predicted to (Takeley, UK) using specific primers flanking exonic regions of IRE1. sensitize GBM cells to chemotherapies such as TMZ. This could for The presence of IRE1alpha mutation was detected using nucleotide instance be achieved by using inhibitors of BiP that were success- sequence alignment software. Tumor in which IRE1 mutation was fully tested in melanoma models (Cerezo et al, 2016). In addition to identified presented classical GBM characteristic with endothelial the direct effect of IRE1 inhibitors on the tumor cells, one might also hyperplasia and MIB1 proliferation index of 15%, and was IDH1 consider their use in combination with current therapies, the most negative, with 5% of OLIG2 and 5% of TP53 positive cells. common of which comprising the combination of radio- and chemotherapy (the latter with the alkylating agent TMZ; Stupp et al, Cloning and site-directed mutagenesis 2005). Provided that about half of GBM patients are resistant to TMZ, their stratification in terms of TMZ sensitization through Selected punctual mutations were introduced on IRE1alpha exonic selective IRE1 inhibition would represent an appealing therapeutic sequence using QuickChange Directed Mutagenesis kit with the alternative. Finally, the link between IRE1 signaling and the most primers described in Appendix Table S6. The wild-type or mutated recent classification of GBM (Louis et al, 2016) as well as whether sequences were then cloned in the multicloning site of the expres- or not IRE1 downstream signals are associated with the classes of sion lentivector pCDH-CMV-MCS-EF1-Puro-copGFP (System bios- GBM with poorer (mesenchymal) or better (pro-neural) prognosis ciences). The presence of only mutations of interest was checked by (Li et al, 2015) remains to be established. a minimum two-X cover sequencing (Beckman Coulter Genomics). Collectively, our work demonstrates for the first time that the uncoupling of XBP1s and RIDD signals downstream of IRE1 impacts Cell culture and treatments on cancer development and points toward an alternative therapeutic avenue coupled with personalized molecular diagnosis for (i) U87MG (ATCC) and U251MG (Sigma, St Louis, MO, USA) cells were decreasing tumor cells’ adaptive properties, (ii) enhancing RNA authenticated as recommended by AACR (http://aacrjournals.org/ catabolic pathways leading to accelerated tumor cell death, and (iii) content/cell-line-authentication-information) and tested for the modulating the tumor stroma through reduced angiogenesis and absence of mycoplasma using MycoAlert (Lonza, Basel, Switzer- increased anti-tumor immunity. These approaches combined with a land) or MycoFluor (Invitrogen, Carlsbad, CA, USA). U87 cells better knowledge of GBM IRE1 signaling characteristics may contri- (ATCC) were grown in DMEM Glutamax (Invitrogen, Carlsbad, CA, bute to develop a new precision medicine tool for GBM treatment. USA) supplemented with 10% FBS. U87 were stably transfected at MOI = 0.3 with pCDH-CMV-MCS-EF1-Puro-copGFP (System biosciences) empty vector (EV), pCDH-CMV-MCS-EF1-Puro-copGFP Materials and Methods containing IRE1alpha wild-type sequence (WT), or mutated sequence (P336L, A414T, S769F, or Q780*). U87 cells were selected Patient samples and primary lines using 2 lg/ml puromycin, and polyclonal populations were tested for GFP expression. Transfections of GBM primary cell lines with The experiments conformed to the principles set out in the WMA IRE1 WT and Q780* were performed using Lipofectamine LTX Declaration of Helsinki and the Department of Health and Human (Thermo Fisher Scientific), according to the manufacturer’s instruc- Services Belmont Report. All tumors were frozen after surgical tions. For microarray experiments, tunicamycin [purchased from resection. These tumors were either clinically or genetically charac- Calbiochem (Merck KGaA, Darmstadt, Germany)] was used at terized in the department of neurosurgery of the Pellegrin Hospital 0.5 lg/ml for 16 h. Actinomycin D was purchased from Sigma- (Bordeaux, France), and informed consent was obtained in accor- Aldrich (St Louis, MO, USA) and used as indicated. For flow cyto- dance with the French legislation or was obtained from the process- metry, antibodies against human CD11b, CD31, and CD45 were ing of biological samples through the Centre de Ressources obtained from BD Biosciences (Le Pont-de-Claix, France). Anti- Biologiques (CRB) Sante´ of Rennes BB-0033-00056. The research miR-17 (miRvana) was purchased from Thermo Fisher Scientific. protocol was conducted under French legal guidelines and fulfilled the requirements of the local institutional ethics committee. GBM Semi-quantitative PCR and Quantitative real-time PCR were classified according to (i) the presence of IDH1, OLIGO2 and TP53 expression and (ii) tumor phenotype (size and form of tumor Total RNA was prepared using the TRIzol reagent (Invitrogen, Carls- cells, hyperplasia, necrosis, proliferation index). Primary GBM lines bad, CA, USA). Semi-quantitative analyses were carried out as were generated as previously described (Avril et al, 2017). Briefly, previously described (Dejeans et al, 2012; Pluquet et al, 2013). PCR fresh tumor tissues were mechanically dissociated using gentle- products were separated on 4% agarose gels. All RNAs were MACS dissociator following the manufacturer’s instructions reverse-transcribed with Maxima Reverse Transcriptase (Thermo

16 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

109 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

Scientific, Waltham, MA, USA), according to manufacturer protocol. mice were injected in immunocompromised nude mice as described All PCRs were performed with a MJ Mini thermal cycler from Bio- (Avril et al, 2017) and above with U87 cells. Mice were daily clini- Rad (Hercules, CA, USA) and qPCR with a StepOnePlusTM Real-Time cally monitored and sacrificed at the first clinical signs. Mouse PCR Systems from Applied Biosystems and the SYBR Green PCR brains were collected, fixed in formaldehyde solution 4%, and Core reagents kit (Bio-Rad). Experiments were performed with at paraffin-embedded for histological analysis after H&E staining. least triplicates for each data point. Each sample was normalized on Tumor burden was compared in the different groups of mice and the basis of its expression of the 18S gene. For quantitative PCR, the analyzed using ImageJ software. Furthermore, vascularization of primer pairs used are described in Appendix Table S7. the tumors (CD31), macrophage infiltration (IBA1), and invasion tumor (vimentin) were monitored using immunohistochemistry Western blotting and immunofluorescence analyses (Appendix Table S8).

Antibodies are described in Appendix Table S8. All IRE1 signaling Statistical analyses analyses were carried out as described previously (Lhomond et al, 2015). Cells grown on 22-mm coverslip were washed with PBS, Data are presented as mean SD or SEM (as indicated). Statistical fixed with 4% paraformaldehyde for 10 min at room temperature, significance (P < 0.05 or less) was determined using a paired or and then blocked with 5% BSA, PBS, 0.1% Triton X-100 for 1 h. ER unpaired t-test or ANOVA as appropriate and performed using was stained using anti-KDEL antibody (Enzo), and overexpressed GraphPad Prism software (GraphPad Software, San Diego, CA, IRE1alpha was stained using anti-IRE1alpha antibody (SantaCruz). USA). Cells were incubated with primary antibodies for 1 h at room temperature, washed with PBS, and incubated for 45 min with Peripheral blood mononuclear cells chemoattraction assay donkey anti-mouse and donkey anti-rabbit antibodies (Invitrogen). To visualize the nucleus, cells were counterstained with 1 lg/ml Peripheral blood mononuclear cells (PBMC) were isolated from 40,6-diamidino-2-phenylindole (DAPI, Sigma). After mounting, cells healthy donors as described previously (Avril et al, 2012). PBMC were analyzed with a SP5 confocal microscope (Leica Microsystems, were washed in DMEM, placed in Boyden chambers (5 × 105 cells/ Mannheim, Germany). chamber in DMEM; Millipore, France) that were placed in DMEM or conditioned medium from cells treated with mirVana miRNA-17 Intracranial injections, tumor size, and blood inhibitor (Thermo Fisher Scientific, Waltham, MA, USA), and then capillary measurements incubated at 37°C for 24 h. The migrated PBMC (under the Boyden chambers) were collected, washed in PBS, and cells were stained for All animal procedures met the European Community Directive monocytes, T, B, and NK cell markers (anti-CD14, anti-CD3, anti- guidelines (Agreement B33-522-2/ No DIR 1322) and were CD19, and anti-CD56, respectively) and analysis by flow cytometry approved by the local ethics committee. Two independent sets of as described below. The relative number of migrated cells was esti- experiments were carried out using 8-week-old male Rag-c2 mice mated by flow cytometry by counting the number of cells per housed in ventilated racks. The protocol used was as previously minute. described (Auf et al, 2010). Cell implantations were at 2 mm lateral to the bregma and 3 mm in depth using seven different sets of cells Tumor migration assay for U87-EV cells, U87-WT cells, U87-S769F cells, U87-Q780* cells, U87-P336L cells, U87-A414T cells, and U87 IRE1-DN cells. Fifteen Parental U251 and U87 cell lines transfected with either empty days post-injection, or at first clinical signs, mice were sacrificed, vector pcDNA 3.1/myc-His B or lenti-pCDH-IRE1 A414T, and subse- brains were frozen, and sliced using a cryostat. Brain sections were quently transfected with siRNA against XBP119 or anti microRNA- stained using H&E staining or anti-vimentin antibodies (Interchim) 17, were washed in DMEM, placed in Boyden chambers (105 cells/ for visualization of tumor masses. Tumor volume was then esti- chamber in DMEM) that were placed in DMEM 20% FBS, and incu- mated by measuring the length (L) and width (W) of each tumor bated at 37°C. After 24 h, Boyden chambers were washed in PBS and was calculated using the following formula (L × W2 × 0.5). and cells were fixed in PBS 0.5% paraformaldehyde. Non-migrated CD31-positive vessels were numerated after immunohistologic cells inside the chambers were removed, and cells were then stained staining of the vascular bed using rat antibodies against CD31 with Giemsa (RAL Diagnostics, Martillac, France). After washes in (PharMingen) and fluorescent secondary antibodies (Interchim). PBS, pictures of five different fields were taken. Migration was given Imaging was carried out using a Axioplan 2 epifluorescence micro- by the mean of number of migrated cells observed per field. For scope (Zeiss) equipped with a digital camera Axiocam (Zeiss). GBM primary cell migration, Boyden chambers were previously Blood vessels were quantified by two independent investigators coated with 0.1% collagen solution (Sigma-Aldrich). using a blinded approach. Vessel number was measured in 12–20 images per condition using ImageJ software. This quantification FACS analyses was made three times for each image, and three vessel sizes (surface) were reported: between 100 pixel² and 500 pixels², more than Glioblastoma multiform specimens were dissociated using the 500 pixel², or more than 5,000 pixel² (1 pixel = 0.67 lm). The aver- gentleMACS dissociator (Miltenyi Biotec, Paris, France) according to age of vessel number of each size was calculated per brain. Experi- manufacturer’s recommendations, and cells were directly used for ments were repeated at least on five Rag-c2 mice for each condition flow cytometry analysis. Cells were washed in PBS 2% FBS and (up to 15). For GBM primary cell-line implantation, 5 × 105 cells/ incubated with saturating concentrations of human

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 17 of 19

110 Published online: January 8, 2018 EMBO Molecular Medicine IRE1 signaling in GBM Stéphanie Lhomond et al

France. M. McMahon was funded by an ARED PhD scholarship from the Région The paper explained Bretagne. This work was carried out under the auspices of REACT (REnnes Problem brAin Cancer Team). Glioblastoma multiform (GBM) is the most lethal primary brain cancer with an overall survival of 15 months and no effective treatment. The Author contributions 1 endoplasmic reticulum stress sensor IRE contributes to GBM progres- SL, ND, NP-L, KB, OPl, and EC carried out experiments on U87 or U251 cells sion, impacting tissue invasion and tumor vascularization. IRE1 is an and studied the mouse in vivo data. TA, DD, GJ, and JO contributed the work RNase that signals by catalyzing the splicing of the mRNA encoding the transcription factor XBP1 and by regulating the stability of certain on primary GBM lines. MMc, KS, MMa, HB, and AS contributed the work on miRNAs and mRNAs through a process known as regulated IRE1- microRNA. KV, OPa, ML, and AC performed all the bioinformatics analyses. RP dependent decay (RIDD). and FJ were in charge of the animal experiments. P-JLR, EV, VQ, AE, and JM contributed the GBMmark cohort. JBP provided the IRE1 RNase inhibitor. CH Results contributed unpublished information and reagents and helped mature the We found that signaling from IRE1 RNase domain defined specific manuscript. SL, ND, OPl, TA, and JO interpreted the data. EC conceived and expression signatures that were relevant in human GBM. This allowed us to demonstrate the antagonistic roles of XBP1 mRNA splicing and realized the study. AC, AS, CH, TA, and EC wrote the manuscript. RIDD on tumor outcomes. Conflict of interest Impact The authors declare that they have no conflict of interest. This study provides the first demonstration of a dual role of IRE1 downstream signaling in cancer and opens a new therapeutic window to impair tumor progression and/or enhance sensitivity to current treatments. References

Anton K, Baehring JM, Mayer T (2007) Glioblastoma multiforme: overview of current treatment and future perspectives. Hematol Oncol Clin North Am immunoglobulins and fluorescent-labelled primary antibodies for 26: 825 – 853 30 min at 4°C. Cells were then washed with PBS 2% FBS and Auf G, Jabouille A, Guerit S, Pineau R, Delugin M, Bouchecareilh M, Magnin N, analyzed by flow cytometry using a FACSCanto II flow cytometer Favereaux A, Maitre M, Gaiser T et al (2010) Inositol-requiring enzyme (BD Biosciences). The population of interest was gating according to 1alpha is a key regulator of angiogenesis and invasion in malignant its FSC/SSC criteria. The dead cell population was excluded using 7- glioma. Proc Natl Acad Sci USA 107: 15553 – 15558 amino-actinomycin D (7AAD) staining (BD Biosciences). Data were Auf G, Jabouille A, Delugin M, Guerit S, Pineau R, North S, Platonova N, analyzed with the FACSDiva (BD Biosciences). Maitre M, Favereaux A, Vajkoczy P et al (2013) High epiregulin expression in human U87 glioma cells relies on IRE1alpha and promotes autocrine Data availability growth through EGF receptor. BMC Cancer 13: 597 Avril T, Vauleon E, Hamlat A, Saikali S, Etcheverry A, Delmas C, Diabira S, The data from this publication have been deposited to the Array- Mosser J, Quillien V (2012) Human glioblastoma stem-like cells are more Express database (https://www.ebi.ac.uk/; tumor transcriptomes sensitive to allogeneic NK and T cell-mediated killing compared with GBMmark) with the accession number E-MTAB-6326 and GEO serum-cultured glioblastoma cells. Brain Pathol 22: 159 – 174 database (https://www.ncbi.nlm.nih.gov/geo/; cell-line transcrip- Avril T, Etcheverry A, Pineau R, Obacz J, Jegou G, Jouan F, Le Reste PJ, Hatami tomes) with the accession number GSE107859. M, Colen RR, Carlson BL et al (2017)CD90 expression controls migration and predicts dasatinib response in glioblastoma. Clin Cancer Res 23: Expanded View for this article is available online. 7360 – 7374 Calfon M, Zeng H, Urano F, Till JH, Hubbard SR, Harding HP, Clark SG, Ron D Acknowledgements (2002) IRE1 couples endoplasmic reticulum load to secretory capacity by We thank Drs S. Manié (Lyon) and R. Pedeux (Rennes) for critical reading of processing the XBP-1 mRNA. Nature 415: 92 – 96 the manuscript, the CRB Santé of Rennes BB-0033-00056 for patient samples, Cancer Genome Atlas Research Network (2008) Comprehensive genomic the BIOSIT High Precision histopathology platform H2P2 (A. Fautrel and P. characterization defines human glioblastoma genes and core pathways. Bellaud; https://biosit.univ-rennes1.fr/?q=en/corefacilities/02/05) for their Nature 455: 1061 – 1068 remarkable work, Biotrial (https://www.biotrial.com/) for help with some Cerezo M, Lehraiki A, Millet A, Rouaud F, Plaisant M, Jaune E, Botton T, Ronco animal models and R. Fronzes (IECB, Bordeaux) for help with IRE13D models. C, Abbe P, Amdouni H et al (2016) Compounds triggering ER Stress exert Affymetrix microarrays were processed in the Microarray Core Facility of the anti-melanoma effects and overcome BRAF inhibitor resistance. Cancer Institute in Regenerative Medicine and Biotherapy, CHU-INSERM-UM Mont- Cell 29: 805 – 819 pellier (http://irmb.chu-montpellier.fr/). This work was funded by grants from Cerrato JA, Yung WK, Liu TJ (2001) Introduction of mutant p53 into a wild- Institut National du Cancer (INCa; PLBIO: 2017-148, PLBIO: 2015-111, type p53-expressing glioma cell line confers sensitivity to Ad-p53-induced INCA_7981), La Ligue Contre le Cancer (Comité des Landes, LARGE project), and apoptosis. Neuro Oncol 3: 113 – 122 PHC Maimonide to EC; EU H2020 MSCA ITN-675448 (TRAINERS) and MSCA Chen X, Iliopoulos D, Zhang Q, Tang Q, Greenblatt MB, Hatziapostolou M, Lim RISE-734749 (INSPIRED) to EC and AS and from Région Bretagne «AAP CRITT E, Tam WL, Ni M, Chen Y et al (2014) XBP1 promotes triple-negative breast santé 2013» to VQ. SL was funded by a PhD scholarship from the French cancer by controlling the HIF1alpha pathway. Nature 508: 103 – 107 government and by a scholarship from the Fondation pour la Recherche Médi- Chevet E, Hetz C, Samali A (2015) Endoplasmic reticulum stress-activated cell cale. ND was funded by a post-doctoral fellowship from the Fondation de reprogramming in oncogenesis. Cancer Discov 5: 586 – 597

18 of 19 EMBO Molecular Medicine 10:e7929 | 2018 ª 2018 The Authors

111 Published online: January 8, 2018 Stéphanie Lhomond et al IRE1 signaling in GBM EMBO Molecular Medicine

Dejeans N, Pluquet O, Lhomond S, Grise F, Bouchecareilh M, Juin A, 2016 World Health Organization classification of tumors of the central Meynard-Cadars M, Bidaud-Meynard A, Gentil C, Moreau V et al (2012) nervous system: a summary. Acta Neuropathol 131: 803 – 820 Autocrine control of glioma cells adhesion and migration through Lu Y, Liang FX, Wang X (2014) A synthetic biology approach identifies the IRE1alpha-mediated cleavage of SPARC mRNA. J Cell Sci 125: mammalian UPR RNA ligase RtcB. Mol Cell 55: 758 – 770 4278 – 4287 Maurel M, Chevet E, Tavernier J, Gerlo S (2014) Getting RIDD of RNA: IRE1 in Dejeans N, Manie S, Hetz C, Bard F, Hupp T, Agostinis P, Samali A, Chevet E cell fate regulation. Trends Biochem Sci 39: 245 – 254 (2014) Addicted to secrete – novel concepts and targets in cancer therapy. Obacz J, Avril T, Le Reste PJ, Urra H, Quillien V, Hetz C, Chevet E (2017) Trends Mol Med 20: 242 – 250 Endoplasmic reticulum proteostasis in glioblastoma – from Drogat B, Auguste P, Nguyen DT, Bouchecareilh M, Pineau R, Nalbantoglu J, molecular mechanisms to therapeutic perspectives. Sci Signal Kaufman RJ, Chevet E, Bikfalvi A, Moenner M (2007) IRE1 signaling is 10: eaal2323 essential for ischemia-induced vascular endothelial growth factor-A Parsons DW, Jones S, Zhang X, Lin JC, Leary RJ, Angenendt P, Mankoo P, expression and contributes to angiogenesis and tumor growth in vivo. Carter H, Siu IM, Gallia GL et al (2008) An integrated genomic analysis of Cancer Res 67: 6700 – 6707 human glioblastoma multiforme. Science 321: 1807 – 1812 Genovese G, Ergun A, Shukla SA, Campos B, Hanna J, Ghosh P, Quayle SN, Rai Pluquet O, Dejeans N, Bouchecareilh M, Lhomond S, Pineau R, Higa A, K, Colla S, Ying H et al (2012) microRNA regulatory network inference Delugin M, Combe C, Loriot S, Cubel G et al (2013) Posttranscriptional identifies miR-34a as a novel regulator of TGF-beta signaling in regulation of PER1 underlies the oncogenic function of IREalpha. Cancer glioblastoma. Cancer Discov 2: 736 – 749 Res 73: 4732 – 4743 Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, Bignell G, Davies Ray A, Zhang S, Rentas C, Caldwell KA, Caldwell GA (2014) RTCB-1 mediates H, Teague J, Butler A, Stevens C et al (2007) Patterns of somatic mutation neuroprotection via XBP-1 mRNA splicing in the unfolded protein in human cancer genomes. Nature 446: 153 – 158 response pathway. J Neurosci 34: 16076 – 16085 Han D, Lerner AG, Vande Walle L, Upton J-P, Xu W, Hagen A, Backes BJ, Oakes SA, Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJB, Papa FR (2009)IRE1alpha kinase activation modes control alternate Belanger K, Brandes AA, Marosi C, Bogdahn U et al (2005) Radiotherapy endoribonuclease outputs to determine divergent cell fates. Cell 138: 562 – 575 plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Hetz C (2012) The unfolded protein response: controlling cell fate decisions Med 352: 987 – 996 under ER stress and beyond. Nat Rev Mol Cell Biol 13: 89 – 102 Upton JP, Wang L, Han D, Wang ES, Huskey NE, Lim L, Truitt M, McManus Hetz C, Chevet E, Harding HP (2013) Targeting the unfolded protein response MT, Ruggero D, Goga A et al (2012) IRE1alpha cleaves select microRNAs in disease. Nat Rev Drug Discov 12: 703 – 719 during ER stress to derepress translation of proapoptotic Caspase-2. Hetz C, Chevet E, Oakes SA (2015) Proteostasis control by the unfolded Science 338: 818 – 822 protein response. Nat Cell Biol 17: 829 – 838 Urano F, Bertolotti A, Ron D (2000) IRE1 and efferent signaling from the Jabouille A, Delugin M, Pineau R, Dubrac A, Soulet F, Lhomond S, Pallares- endoplasmic reticulum. J Cell Sci 113(Pt 21): 3697 – 3702 Lupon N, Prats H, Bikfalvi A, Chevet E et al (2015) Glioblastoma invasion Urra H, Dufey E, Avril T, Chevet E, Hetz C (2016) Endoplasmic reticulum and cooption depend on IRE1alpha endoribonuclease activity. Oncotarget stress and the hallmarks of cancer. Trends Cancer 2: 252 – 262 6: 24922 – 24934 Volkmann K, Lucas JL, Vuga D, Wang X, Brumm D, Stiles C, Kriebel D, Der- Jurkin J, Henkel T, Nielsen AF, Minnich M, Popow J, Kaufmann T, Heindl K, Sarkissian A, Krishnan K, Schweitzer C et al (2011) Potent and selective Hoffmann T, Busslinger M, Martinez J (2014) The mammalian tRNA ligase inhibitors of the inositol-requiring enzyme 1 endoribonuclease. J Biol complex mediates splicing of XBP1 mRNA and controls antibody secretion Chem 286: 12743 – 12755 in plasma cells. EMBO J 33: 2922 – 2936 Walter P, Ron D (2011) The unfolded protein response: from stress pathway Kosmaczewski SG, Edwards TJ, Han SM, Eckwahl MJ, Meyer BI, Peach S, to homeostatic regulation. Science 334: 1081 – 1086 Hesselberth JR, Wolin SL, Hammarlund M (2014) The RtcB RNA ligase is Xue Z, He Y, Ye K, Gu Z, Mao Y, Qi L (2011) A conserved structural an essential component of the metazoan unfolded protein response. determinant located at the interdomain region of mammalian inositol- EMBO Rep 15: 1278 – 1285 requiring enzyme 1alpha. J Biol Chem 286: 30859 – 30866 Lhomond S, Pallares N, Barroso K, Schmit K, Dejeans N, Fazli H, Taouji S, Yoshida H, Matsui T, Yamamoto A, Okada T, Mori K (2001) XBP1 mRNA is Patterson JB, Chevet E (2015) Adaptation of the secretory pathway in induced by ATF6 and spliced by IRE1 in response to ER stress to produce cancer through IRE1 signaling. Methods Mol Biol 1292: 177 – 194 a highly active transcription factor. Cell 107: 881 – 891 Li R, Li H, Yan W, Yang P, Bao Z, Zhang C, Jiang T, You Y (2015) Genetic and clinical characteristics of primary and secondary glioblastoma is License: This is an open access article under the associated with differential molecular subtype distribution. Oncotarget 6: terms of the Creative Commons Attribution 4.0 7318 – 7324 License, which permits use, distribution and reproduc- Louis DN, Perry A, Reifenberger G, von Deimling A, Figarella-Branger D, tion in any medium, provided the original work is Cavenee WK, Ohgaki H, Wiestler OD, Kleihues P, Ellison DW (2016) The properly cited.

ª 2018 The Authors EMBO Molecular Medicine 10:e7929 | 2018 19 of 19

112 APPENDIX - Supplementary Files

Table of Content %!! $"'$"# $ # %!! $"'# 2:;E>1 .:@> 2:;E>1 .:@> 2:;E>1 .:@> 2:;E>1 .:@> 2:;E>1 .:@> 2:;E>1 .:@> 2:;E>1 .:@> 2:;E>1 .:@> %!! $"' %" # %!! $"'" # %!! $"' %"# $B@NK>1 .:@> $B@NK>1 .:@> $B@NK>1 .:@> $B@NK>1 .:@> $B@NK>1 .:@> $B@NK>1 .:@> $B@NK>1 .:@> $B@NK>1 .:@>

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`L BGLMKNILP>K>O:EB=:M>=NLBG@02 J.!0:G=02 .!0:@:BGLM`HK`LI>KL H? '0# F0, M:K@>ML .#0 :G= 6 . HK :@:BGLM F0, % ."& + :G= GHG F>LL>G@>K0, 1KB;HLHF:E0, 1)0, 3LIEB<>HLHF:E0, GHMLAHPG BHMBGRE:M>=LBG@E>LMK:G=<", P:LMA>GIK>I:K>=:<52 + '-,DBMF:GN:E 0>O" 0>O" ?KHF G@ MHM:E 0, :G= 52%>G> ??RF>MKBQF:GN:E0>O., 0>O $HEEHPBG@?K:@F>GM:MBHG:G=M>KFBG:EE:;>EBG@ N@H?LBG@E>LMK:G=<", P>K> AR;KB=BS>= ?HK A :M W! HG ??RF>MKBQ %>G>!ABIL %>G>!ABIL P>K> P:LA>= :G= LM:BG>=BGMA> ??RF>MKBQ$ENB=B=NLBG@MA> ??RF>MKBQ

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

115 .10MHMA>BK>G= ;:L>=HGMA>LMKLNL!RP>K>:;L>GM HK=HPGK>@NE:M>=;:L>=HGMA>G=KLNL"LI>ER P>K> K>=:LIHM>GMB:EM:K@>MLH?'0# ! ] $HK <>EENE:K :G:ERL>L MHM:E 0, P:L >QMK:= NLBG@ MA> ,NH1IBG0, '' MHM:E0, BLHE:MBHGDBM+:K>R ,:@>E "VK>G %>KF:GR ?KHF3#4 3", 352 31$ 3/ <>EEL 1:FIE>LP>K>MA>G:G:ERL>=NLBG@ ??RF>MKBQ :KK:RL DBM &% 3 :M MA> .E:M>?HKF ^.N<>L U ", 2&"_ BG +HGMI>EEB>K 2A> :KK:R BGM>GLBMR LB@G:EL P>K> :G:ERS>= ?NLBG@ MA> EB;K:KB>L H? %>G> 0+ " LH?MP:K> .K> IKH<>LLBG@P:LI>K?HKF>=PBMAMA>0+ F>MAH=:G=MA>GHKF:EBS:MBHGP:LI>K?HKF>=PBMA MA> /N:GMBE> F>MAH= $KHF MA> GHKF:EBS>= O:EN>L MA> !H>??BGM H? 4:KB:MBHG !4 P:L >LMBF:M>= ?HK >: 'G HK=>K MH =>KBO> =B??>K>GMB:EER >QIK>LL>= IKH;>L MA> <>EEL MK:GL?>=PBMAMA>>FIMRO>IK>L>GM>=MA>@HKR:G=LM:MBLMB<:EL>E>

P:L;:L>=HGMA>L<:E>=!4F>MAH=PBMAMA>?HEEHPBG@KB:?HE=TT:G=!4L<:E>= 'GHK=>KMHI>K?HKF?NGL>MLH?=B??>K>GMB:EER>QIK>LL>= @>G>L =>KBO>= ?KHF MA> :?HK>F>GMBHG>= LM:MBLMB<:E >QIEHBM>= MA> $'&"'$&!)IE:M?HKF!A:MSBBH:GGHN +HNML>EHL>M:E +HNML>EHL>M:E .BE:EBL!A:MSBBH:GGHN PABK?HKFLLM:MBLMB<:E:G=G>MPHKD:G:ERLBLHGO:KBHNL ;BHEH@B<:EAB>K:KL%>G>-GMHEH@R%- LA;NKG>K>M:E 0>: I:MAP:RL!KH?M>M:E MA>&NF:G.A>GHMRI>-GMHEH@R&.-)HAE>K>M:E :G= +%'+:FF:EB:G-GMHEH@R#IIB@>M:E :BFBG@MH=>M>K>= ;BHEH@B<:E IKH<>LL>L :G= MA> K>LI> =KBO>K @>G>L EBGDBG@ MA>L> IKH<>LL>L 2A> $'$&"'%$&!)IBI>EBG>BL:O:BE:;E>HGEBG>:MMA>P>;LBM>AMMIL;BHBG?HFBG>K K:KKBG@ P:L I>K?HKF>= HG MA> NGBHG H? MA> =B??>K>GMB:EER >QIK>LL>= IKH;>L ?KHF :EE MA> L:FIE>LNLBG@M>EBGD:@>F>MAH=:G==BLM:G<>F>MKB< $HK MNFHK :G:ERL>L : EH<:E GML MK>:M>= PBMA K:=BHMA>K:IR :G= GM:=CNO:GMM>FHSHEHFB=>BG:<PBMAMA>LM:G=:K=H?<:K>% +F:KDP:L K>MKHLI>ER K>= :G= NL>= ?HK MK:GL :G:ERLBL % + LI>GL P>K> H;M:BG>=:?M>KBG?HKF>=GM?KHFI:MB>GML:=FBMM>=MHMA>G>NKHLNK@>KR=>I:KMF>GM:M 0>GG>L 3GBO>KLBMR &HLIBM:E ?HK LNK@B<:E K>L> PBMA MA> EH<:E >MAB< > 2NFHKLNL>=BGMABLLMN=RP>K>ABLMHEH@B<:EER=B:@GHL>=:L@K:=>'4:LMKH 5&- KB: 2NFHK L:FIE>L P>K> LG:I ?KHS>G BFF>=B:M>ER :?M>K K>L>L IK>L>GM>= :M E>:LM H? MNFHK <>EEL 2A> >QM>GM H? LNK@>KR P:L >O:EN:M>=PBMA:G>GA:G<>=F:@G>MBLHG:G<>BF:@BG@+0'I>K?HKF>=PBMABGAHNKL :?M>KMA>K>L>=PBMAMA>,NH1IBG0, '')BM+:K>R ,:@>E &H>K=M $K:G<> 0, BGM>@KBMR0, 'GM>@KBMR,NF;>K[P:L=PBMA:G @BE>GM ;BH:G:ERS>K @BE>GM 2>L *>L 3EBL $K:G<> %>G> >QIK>LLBHG IKH?BEBG@ P:L

116 <:KKB>= HNM PBMA MA> @BE>GM PAHE> ANF:G @>GHF> Q) FBGM 2>L 2HM:E 0, P:L >QMK:= E:;>EE>= :G= AR;KB=BS>=:<DBM F:GN?:K`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

 118 

80&)52*)1)6 )1)6 #)50 ;75%')//8/%50%75-;25+%1-=%7-21 )+5%(%7-212*7,));75%')//8/%50%75-; "<1()'%117)5%'7-216 ( 2//%+)1*250%7-21 2//%+)1()+5%(%7-21 21-17)+5-10)0&5%1)-17)5%'7-216 )+$(%2-%) 17)+5-1')//685*%')-17)5%'7-216 0-+ "&&.&+'-+%0 /%67-'*-&5)*250%7-21 %0-1-1-17)5%'7-216 ( "-+1%/-1+&<5)7-12-'%'-( !)+8/%7-212*#6-+1%/-1+

%$(& %675-1!6-+1%//-1+3%7,:%<9-% %1( 

+(,!. -%) %/3,%46-+1%//-1+)9)176 

, 20)267%6-6 

200213%7,:%<2**-&5-1'/27*250%7-21 )//685*%')-17)5%'7-216%77,)9%6'8/%5:%// 

)'"),-,% %6-+-1-17)5%'7-216 119 17)5/)8.-1 352')66-1+ 17)5*)5216-+1%/-1+ 1(2620%/$%'82/%53%7,:%< 17-+)1 5)6)17%7-212/(-1+%66)0&/<%1(3)37-()/2%(-1+2* '/%66 #5%**-'.-1+%1(352')66-1+2*)1(2620%/#! ''.("1,-"' 8'/)27-()&-1(-1+(20%-1/)8'-1)5-',5)3)%7'217%-1-1+5)')3725 6-+1%/-1+3%7,:%<6 -/)%'-(%1(6%/70)7%&2/-60

(, "<17,)6-62*7,)&-/)%'-(6%1(&-/)6%/76 ')# "- )&%, &%*)*+)-"% &%*%!,(! "<17,)6-62* 5267%+/%(-16%1(#,520&2;%1)6 !)7-12-(0)7%&2/-60%1(#5%163257 )# (, )&%,' /%-'% "- $-7%0-1 0)7%&2/-60 , , !)+8/%7-212*168/-1/-.)75%163257%1(837%.) 26775%16/%7-21%/3527)-102(-*-'%7-21 ')# *+)-"%( "- )&% 63%5%+-1) /-1.)(+/<'26

 

& 1)FE8:H?4G87:8A8F54F87BAGE4AF6E

120 & %

AAK1 C12orf73 DAGLB G6PD KCNAB2 MAB21L1 OBSCN RAB3D SBF2 TAB1 UBE2D3 ABCA1 C15orf41 DBN1 GALK1 KCNH3 MACF1 OCM2 RAB3GAP2 SBF2-AS1 TACC1 UBE2F ABCA3 C16orf89 DCBLD2 GALNT3 KCNH4 MAD1L1 OLFML2B RAB5A SCARB2 TAF2 UBE2G1 ABCB4 C17orf53 DCLK1 GANAB KCNJ5 MAEA OPHN1 RABGEF1 SCARNA13 TAF9B UBE2Z ABCB9 C1orf159 DCTN2 GATSL3 KCNJ9 MAGI2-AS3 ORAOV1 RAD17 SCEL-AS1 TAOK1 UBE4A ABI2 C1QTNF9B DCTN5 GBE1 KCNS1 MAN2A1 OSBPL1A RAD50 SCFD1 TAOK3 UBR2 AC005550.5 C2 DDB2 GCGR KDM8 MANSC1 OSBPL2 RAI14 SCGN TARS UBR5 AC007271.3 C21orf67 DDX3Y GGA2 KIAA0195 MAOA OSMR RALGAPA2 SCLY TAT UCKL1 AC009158.1 C2CD3 DDX5 GHITM KIAA0355 MAP2K2 OSTM1 RANBP2 SEC23A TBC1D12 UFM1 AC010880.1 C3 DDX60L GIGYF2 KIAA1033 MAP3K10 OTUD4 RARRES1 SEC23IP TBC1D9 UGDH AC093495.4 C5orf42 DGCR14 GK5 KIAA1109 MAP3K14 OTUD5 RASA4 SEC31A TBL1XR1 UGGT2 AC096559.2 C7orf43 DGKQ GLB1 KIAA1191 MAP3K6 OTUD7A RB1 SEC63 TCOF1 UHRF2 ACAN C9orf117 DHX9 GLG1 KIF1C MAP4K3 OXCT1 RBBP9 SEL1L3 TCP11L1 ULK3 ACBD5 C9orf173 DICER1 GMFB KIF20A MAPK1 OXR1 RBCK1 SELT TDRD9 ULK4P1 ACLY C9orf3 DIRC2 GNAQ KIF21A MAPK10 PABPC1 RBL1 SEMA4D TECPR1 UNC13A ACO1 C9orf62 DKFZP434K028 GNN KIF21B MAPK1IP1L PAIP1 RBM33 SENP1 TECR UNC80 ACTG1 CA7 DLEU2 GNPTAB KIF2A MAPKAP1 PAK2 RBM39 SENP6 TENC1 UNK ACVR1B CACNA2D4 DLG1 GOLGA8A KIF5B MARCH2 PAK4 RCN1 SEPT10 TFDP1 UPB1 ACYP2 CADPS DLG3 GPNMB KIF6 MARCH6 PAM REPS1 SEPT11 TFE3 URB1 ADAM23 CALM2 DLGAP1 GPR107 KIR2DS3 MARK4 PAN3 RFPL2 SEPT2 TGFBI UROD ADAM9 CALM3 DMXL2 GPR39 KLF7 MARS PAPD4 RFX3 SETX THADA USMG5 ADAMTS12 CAMK2D DNAH14 GPR97 KLHL20 MASP2 PAPD5 RFX7 SF3B1 THAP3 USP24 ADAMTSL1 CANX DNAH17 GRAMD1B KLHL23 MAST3 PAPOLA RGL2 SH3BGRL THAP5 USP34 ADD1 CAPN1 DNAH5 GRIK3 KLHL24 MATR3 PARK7 RGPD6 SH3BGRL2 THOC5 USP37 ADRM1 CAPRIN2 DNAJB4 GRM4 KLHL3 MBNL2 PARN RGS14 SH3RF2 THSD4 USP3-AS1 AHI1 CARKD DNAJC13 GRN KMT2A MBNL3 PARP4 RIF1 SHANK3 TIA1 USP4 AIM1L CASC1 DNASE1L1 GRXCR1 KMT2E MBTPS2 PCBP2 RNF123 SHF TIAF1 USP43 AK9 CASP3 DNM1 GSK3B KPNA2 MCAT PCNX RNF20 SIGLEC1 TIPRL USP45 AKR1A1 CASP7 DNM1P47 GSR KPNA5 MCOLN1 PCNXL2 RNF217 SIN3B TLK1 USP9X AKR1C1 CAST DNMT1 GTF2I KPTN MED1 PCNXL4 RNGTT SKIV2L2 TLL2 UVRAG ALDH1L2 CAV2 DPM1 GTF2IP1 KRT10 MED13 PCSK5 ROBO1 SKOR1 TLN2 VAMP7 ALDH3A2 CBR3-AS1 DPP4 GUCA1A KTN1 MED15 PCYOX1 ROCK1 SLAIN2 TM9SF4 VEPH1 AMMECR1 CBWD1 DPY19L3 GUCY2F LAMA1 MED23 PDCD10 ROPN1 SLC16A1 TMED10 VIM AMY1B CBWD5 DPYD GUSBP9 LAMA4 MEG9 PDCD2 RP11-11N9.4 SLC22A31 TMEFF1 VIPR2 ANAPC1 CBX3 DPYSL2 H2AFZ LAMB4 MFSD11 PDCD6IP RP11-15B24.5 SLC25A45 TMEM131 VMP1 ANAPC16 CBX7 DRAM1 HCG21 LAMTOR1 MGAM PDE8A RP11-174G17 SLC35A5 TMEM161B VPS13A ANK3 CCDC103 DSCAM HDAC6 LARP1 MGMT PDIA5 RP11-18B16.2 SLC35B1 TMEM177 VPS13C ANKRD13B CCDC132 DSE HDAC9 LARP7 MICAL1 PDILT RP11-196D18.1 SLC37A2 TMEM181 VPS41 ANKRD28 CCDC149 DSG2 HECTD4 LATS2 MICB PDLIM2 RP11-216N14.8 SLC39A14 TMEM209 VTI1A ANKRD36 CCDC18 DST HELB LBR MIR1302-2 PDS5A RP11-217E22.5 SLC41A3 TMEM234 VWA5A ANKRD36B CCDC37 DYNC1H1 HERC2 LCLAT1 MIR143HG PER2 RP11-244K5.4 SLC6A8 TMEM243 VWA8 ANKS6 CCDC47 ECHDC1 HERC2P2 LDHA MLYCD PFKP RP11-348F1.3 SLC9A9 TMEM30A WDR3 ANXA5 CCDC50 EDRF1 HERC2P3 LDHB MME PHC3 RP11-436F23.1 SLC9C1 TMEM51 WDR44 ANXA6 CCND3 EHBP1 HHEX LEPRE1 MMP2 PHF20L1 RP11-552M14.1 SLK TMEM59 WDR46 ANXA8 CCNL1 EHBP1L1 HIF1A LGALS1 MON1A PHKB RP11-767N15.1 SLMAP TMEM87B WDR6 ANXA8L2 CCPG1 EIF3E HILS1 LGMN MORF4L2 PI4K2B RP11-782C8.5 SLTM TMEM88B WFDC3 AOC1 CCT3 EIF4E3 HIPK1 LGR4 MOSPD1 PIGT RP4-613B23.1 SMARCA1 TMOD3 WHSC1L1 AP1S2 CCT6P3 EIF4EBP2 HK2 LILRA2 MPDZ PIK3C3 RP5-1180C18.1 SMARCC1 TMPRSS15 WNT5A AP3B1 CCZ1B EIF5A HLA-A LINC00035 MPP6 PIK3CB RP6-74O6.3 SMARCD3 TMPRSS6 WWP1 AP3M2 CD200R1 ELK3 HLA-DPB1 LINC00317 MPRIP PIKFYVE RPL24 SMC4 TMTC3 XPNPEP1 APCDD1L-AS1 CD38 ELMOD2 HLA-DRA LINC00461 MRPL3 PITRM1 RPL3 SMC6 TNC XPO1 APP CD44 ELOVL5 HLA-DRB1 LINC00559 MSH3 PITX2 RPL31 SMCHD1 TNFRSF13B XRCC3 APPBP2 CD58 ENPP1 HLA-DRB3 LINC00593 MSI1 PIWIL3 RPL32 SMIM24 TNFRSF14 XRCC5 APPL2 CDC42BPA ENPP6 HLA-H LINC00636 MSMO1 PKD1 RPL38 SMURF1 TNKS2 YIPF6 ARFGEF2 CDC5L EPAS1 HMCN1 LINC00649 MTA3 PKD2 RPL41 SNORA71E TNN YME1L1 ARHGAP18 CDC73 EPHA6 HMCN2 LINC00893 MTAP PKHD1 RPN2 SNORD50A TNNC2 ZBED1 ARHGAP26 CDCA5 EPHB6 HNRNPA1P10 LINC00959 MTHFD1 PKIA RPS13 SNX13 TNNT2 ZBTB22 ARHGAP39 CDCA7 EPT1 HNRNPK LIPA MTHFD2 PLA2G2A RRH SNX29 TNNT3 ZBTB38 ARHGAP8 CDH1 EPX HNRNPL LITAF MTMR2 PLA2G4E RRM1 SNX3 TNXB ZDHHC20 ARHGEF11 CDH22 ERBB2IP HNRNPUL1 LLGL1 MTMR7 PLAA RRN3 SORBS2 TOLLIP ZEB1 ARHGEF7 CDK11A ERCC6 HOXD10 LMCD1-AS1 MYH10 PLAT RSBN1L SPAG1 TOP1 ZFAND6 ARID2 CDK12 ERGIC3 HSD17B7 LOC100132891 MYH16 PLAU RSRC2 SPARC TOP2B ZFC3H1 ARIH2 CDK19 ERICH6 HSP90B1 LOC100422737 MYH3 PLCD1 RSRP1 SPARCL1 TOR1AIP1 ZFP64 ARL2BP CDK4 ERLIN2 HSPA1B LOC100996291 MYLK PLCE1 RTTN SPATA21 TOX ZFYVE1 ARMC8 CDON ESRP2 HSPA6 LOC101060483 MYO10 PLOD2 RUNX1T1 SPATA31A5 TPPP3 ZHX3 ARVCF CEL ETFA HSPA8 LOC101927641 MYO15A PLP1 RYR2 SPCS3 TPR ZIC1 ASMTL-AS1 CELSR2 EXOSC10 HSPB1 LOC101927768 MYO15B PLS3 SPEF2 TRAF3IP2 ZNF131 ASXL2 CEP104 EXT2 HSPB11 LOC101927843 MYO3B PMP22 SPIDR TRAM1 ZNF215 ATAD2B CEP112 EXTL2 HSPG2 LOC101927902 MYO9B PMVK SPIRE1 TRAPPC8 ZNF22 ATAD3A CEP63 FABP3 HUWE1 LOC101928495 MYT1L PNCK SPOPL TRIM24 ZNF227 ATAD3C CERS2 FADS1 IDH3B LOC101928782 NAA15 PNISR SPRYD7 TROVE2 ZNF23 ATG2B CERS4 FADS2 IFRD1 LOC101928978 NAA40 PNMA2 SPTBN5 TRPS1 ZNF274 ATIC CFC1B FAM114A1 IFRD2 LOC101929653 NAMPT PNPLA8 SQLE TRPV2 ZNF280C ATM CFDP1 FAM118A IFT122 LOC101930595 NBEAL1 POC1B SQSTM1 TSHZ2 ZNF283 ATMIN CHD2 FAM13B IGHG2 LOC283299 NBEAL2 POLL SRBD1 TSPO2 ZNF331 ATP1A1-AS1 CHDH FAM13C IGHMBP2 LOC340515 NBN POM121C SRCIN1 TSTA3 ZNF365 ATP1B3 CHM FAM179B IKBKAP LOC644961 NBPF9 PORCN SREK1 TTC13 ZNF398 ATP1B4 CHMP5 FAM199X IKBKB LOC645202 NCL POU4F3 SRP14 TTC17 ZNF492 ATP6V1B2 CHN1 FAM204A IKZF2 LOC645355 NCOA7 PPIP5K2 SRP72 TTC28 ZNF503-AS1 ATP6V1H CHRD FAM208A IL10RB LOC645513 NCOR1 PPP1CB SRP9 TTC3 ZNF529 ATRN CIDECP FAM86C2P IL17RE LOC650368 NCOR2 PPP1R13B SRPX TTLL5 ZNF555 ATRX CKAP2L FAM89B IL6R LRBA NCSTN PPP2R3C SRPX2 TUBA1A ZNF672 ATXN3 CLASP2 FAM98A IL6ST LRP2 NDEL1 PPP2R5C SRR TUBA1B ZNF93 ATXN7L2 CLASRP FAP IL9R LRPPRC NDRG3 PPP2R5D SRRT TXNDC11 ZNRD1 AWAT2 CLCN5 FARSB ILK LRRC1 NDST2 PPP3CC STAG1 TXNIP ZSWIM8 B2M CLIP2 FAS INTS3 LRRC28 NDUFA3 PPP4R1 STAG2 TXNL1 ZZEF1 B3GAT1 CLN6 FASN INTS6 LRRC37A NDUFA5 PRCC STEAP2 B3GNT2 CLU FASTK IPO5 LRRC37BP1 NDUFA8 PRDM13 STEAP4 B3GNT5 CNBD1 FAT2 ITCH LRRC41 NDUFB8 PRKAA1 STIL BACH1 CNOT6 FBXL3 ITFG1 LRRC45 NEK1 PRKAG1 STK10 BARHL1 CNPY2 FBXL5 ITGAM LRRFIP1 NELL1 PRKAR1B STK3 BBS4 COG6 FBXO27 ITGAV LRRK2 NETO2 PRKAR2A STK39 BBS7 COL11A1 FBXO33 ITGB1BP1 LSM14A NF1 PRKCI STMN3 BBS9 COL11A2 FBXO38 ITGBL1 LTBP1 NFE4 PRKD3 STPG1 BCAP31 COL12A1 FBXW11 ITIH5 LTBP2 NFRKB PRKDC STX2 BCKDHB COL14A1 FCGR2C ITPR2 LUC7L3 NFX1 PRKG2 STXBP2 BCL7C COL15A1 FCHO2 ITPR3 LY9 NGB PRKY STXBP5 BCOR COL18A1 FER1L4 JUP LYRM1 NIPA2 PROM1 SUCO BIRC6 COL18A1-AS2 FERMT2 LYST NIPSNAP1 PROS1 SUFU B2M COL22A1 FKBP15 NKAIN4 PRPS2 SULF1 B3GAT1 COL4A2 FKBP7 NLRP12 PRRC2C SUN2 B3GNT2 COL4A5 FLII NME1 PSAP SVOP B3GNT5 COL5A1 FN1 NNT PSMA7 SYNJ2 BACH1 COL9A2 FOXJ3 NOL8 PSMB10 SYT1 BARHL1 COPB1 FOXP2 NONO PSMB5 SYTL3 BBS4 COPG2 FRAS1 NOP58 PSMD13 BBS7 COPS2 FRMPD2 NOX5 PTCH1 BBS9 COPS7A FSCN3 NPIPA1 PTK2 BCAP31 CORO1C FXR1 NPM1 PTK2B BCKDHB CPB2-AS1 NR2F1-AS1 PTMA BCL7C CPEB2 NRAS PTP4A2 BCOR CPNE3 NRD1 PTPLA BIRC6 CPSF2 NT5C PTPLB BLM CPSF6 NT5DC2 PTPN1 BMS1 CPSF7 NTNG1 PTPN11 BNIP1 CPT1B NUB1 PTPN14 BOD1L1 CREB1 NUCB2 PTPRA BOLA3 CREB3L2 NUDCD1 PTPRK BOP1 CRELD1 NUSAP1 PTPRM BPI CROCCP2 PTPRN BPTF CSNK1G3 PTPRQ BRCA2 CSTF3 PTRF BRD7 CTA-342B11.2 PUM1 BTBD1 CTB-174D11.2 PUM2 BTN1A1 CTD-2335O3.3 PUS3 BZW1 CTNNB1 PUSL1 CTSF PWP2 CTU2 QARS CWC22 CYB5D1 CYC1 CYFIP1 CYP27A1

121 & +" E8:H?4G87 :8A8F 

 122 & %

 123 & )E<@8EFHF879BEF

 124 & +- D)+CE<@8EFHF87

125 +/ - -- 1)GBG 0 - ---- +/ - - - &&) 0 - --- +/ - -- "+ 0 - -- - +/ - -- - - 1% 0 - - - -- +/ ----- - ---- % 0 - -- - +/ --- ------ "% 0 --- -- +/ - ---- - --- "% 0 - --- ---- +/ - --- --

 126 & 4AG<5B7<8FHF87

Antigen Usage Dilution Company Reference CANX WB 1/1000 N/A Chevet et al. EMBOJ (1999) XBP1s IHC 1/200 N/A Lhomond et al. Meth. Mol. Biol. (2015) VIM IHC 1/250 DAKO M0-725 clone B9 11300 IBA1 IHC 1/050 WAKO 019-19741 CD31 IHC 1/050 DIANOVA DIA3105231 IRE1 WB 1/1000 SANTACRUZ H-190

PDGFRB WB 1/1000 SANTACRUZ 958 P53 WB 1/1000 ABCAM DO-1 ACTIN WB 1/1000 ABCAM ab8227 CD11b FACS N/A BD ICRF44 CD45 FACS N/A BD 2D1 KDEL IF 1/250 ABCAM ab12223 CD31 FACS N/A BD WM59

127 $".+ $)$(!)+'-$, 0)+%!&)0 !)+ -# $ (-$!$-$)( )! #. " ( , + *+ , (--$/ )!-# ,$"(-.+ +4J74G4 %94:8 4A7 7<998E8AG<4??L8KCE8FF87 :8A8FJ8E8F8?86G8758GJ88A '4A70-.68??F5LHF6B;BEG54F87BA;<:;BE?BJ"+46G8EF @+' " 8E @+' 8E@+'+!( 6B;BEGGH@BEF8K;<5 4A7 "+4?C;4 8KBAF F8DH8A68F J8E8 6B@C4E87 GB ABE@4? 5E4 8 8A64CFH?4G87 GH@BE !8@4GBKL?

 128 "+ CEBG84 "+4?C;4 CEBG8 4 "+4?C;4 CEBG8 8E4AG< $% :E88A ,) @87<4G87

 129 B9C@+'8KCE8FF 6B;BEG 54F87 BA G;8 1)F F<:A4GHE8& "@@HAB;4AG<5B7<8F $".+ CCEB46; 9BE <78AG<9L

130 F;5HEA8E & 4?? ?4>8 # BGFG8 < $ J<:;G,, CC<:#-8G4? 8A8BAGB?B:LGBB?9BEG;8HA<9<64GABJ?87:854F8 'H6?8<6 6<7F +8F 8=84AF' )?HDH8G( %;B@BA7, E8# H?G# &BHF8 8AB@8 "A9BE@4G<6F& "E89?86G8A, &HA:4??# 4H8E, EBJA % EH7AB & 4@C58?? # 8G 4? -;8 !H@4A );8ABGLC8 (AGB?B:L CEB=86G ?FGBJ4E7FG;8

 131 +4E4$ ,H?9BE;B74@

132 48527 -!!!'". $ $# #)(4235 59" - ;2)26/ * +4-)).+<4)6. - ;2)26/ +4-)).+<3)6. "#3263)) "-98, %") 5 98,&#. - ". #$%" " "$* 0 %!%$ $ 59 #$# ,+64571 448" :8"

#3 "$ 09:#1 $') " "$*$ 0576#1 # $" 0;9#1 $ )"# " 0546#1 $#+# "$) #$ %/#'$ !

) " $" &% <4,48 $#$" "# ) " $") $#$ $""$$"$ "# $" " "$*$ #$ $ $$# ! %/# "$)( "## %4 "$ #0,,1 0" / "$)##1 " &# $"$$ $ #)%/ #$

 

 133

 

 24000 3200 16000

18000 2400 12000

12000 1600 8000

6000 800 4000 0 0 0 HIGH LOW HIGH LOW HIGH LOW 

 

 134 

 

 ! !! 

 

 135 %

 !α 

 ! "-$( 

 

'$+ #/*+#,,$)(#& -$.#$(-#(,$-0 

 

 136 

 

 137

 

 138 

 

 100 100 

75 75 

50 50 25 25

0 0 0 1000 2000 3000 4000 0 500 1000 1500 2000 

 

 139 

 

 140 Chapter 4: Conclusions and Contributions

In this fourth chapter we have established that the IRE1 arm of the UPR plays a major role in GBM development. By dissecting IRE1 RNase activity in human GBM tumours, commonly available and primary GBM cell lines we have revealed the dual role of XBP1s and RIDD in tumour aggressiveness. GBM tumours cluster in high and low IRE1 activity groups and XBP1s conveys tumorigenicity by promoting angiogenesis and macrophage recruitment whilst RIDD dampens angiogenesis and tumour cell migration. Patients bearing tumours with XBP1s+/RIDD- features have a lower survival than those with XBP1s-/RIDD+ features thereby pertaining to potential novel therapeutic avenues. We have thus overseen the completion of the first primary objective by producing a cellular translational GBM model of altered IRE1 activity and by establishing a strong role of IRE1 in GBM pathophysiology.

The author of this thesis contributed to the publication of “Dual IRE1 RNase functions dictate glioblastoma development” the following: Original text for the materials and methods and full experimental work in the production of stable primary lines overexpressing various forms of IRE110 producing figure 8.

10 Special thanks to G. Jegou, Dr T. Avril and Dr. E. Chevet for providing extensive training and guidance during the generation of this GBM model.

141 Chapter 5 IRE1 signalling maintains glioblastoma cells differentiation through the XBP1s/miR-148a-mediated repression of stemness transcription factors

142 Chapter 5: Foreword

In chapter 4 we explored the dual role of IRE1 in GBM carcinogenesis and delved into the details of differential XBP1s versus RIDD signalling. We established how IRE1 acts as a major mediator of GBM tumour formation and angiogenesis as well as how miRNA plays a significant role in disease, all the while establishing an experimental cell model that recapitulates GBM tumours whilst providing an altered IRE1 state. In chapter 5 we shall delve deeper into the role of IRE1 in the maintenance of a specific GBM phenotype. As such we shall analyse properties through which IRE1 may affect the differentiation and reprogramming capabilities of GBM cells, having a profound impact on clinical outcomes as, as established in the introduction, GBM stem properties may be responsible for GBM cell chemotherapy evasion and hence recurrence. We shall establish a signature of genes including transcription factors involved in differentiated cell reprogramming which will be confronted to an IRE1 activity signature established in chapter 4. Thereafter the levels of these transcription factors will be investigated in high and low IRE1 activity tumours, thereby outlining the potential relevance f IRE1 function to their expression. A cellular model of stem-like and differentiated –like phenotype will be established in primary GBM cells, grown in media specific to adherent or neurosphere cell lines. Using the cell models of altered IRE1 activity described in chapter 4 we construct a full genotypic and phenotypic reprogramming landscape by forcing differentiation in the presence or absence of genetic or pharmacological IRE1 inhibition, measuring stem, differentiation and reprogramming markers at the mRNA and protein level as well as observing clonogenic ability and general phenotype. Upon establishing a clear link between IRE1 signalling and GBM cell reprogramming we will witness a further focus on the exact signalling pathway downstream of IRE1 that may govern differentiated phenotype maintenance and will thus uncover the role of XBP1s as an inducer of a specific subset of miRNAs that negatively regulate the transcription factors involved in differentiation and reprogramming deregulating the differentiation maintenance phenotype. As such we will attempt to fulfil the second major objective of this thesis.

143 IRE1 signaling maintains glioblastoma cell differentiation through the XBP1s/miR-148a-mediated repression of stemness transcription factors.

Dimitrios Doultsinos1,2, Mari McMahon1,2,3, Konstantinos Voutetakis4,5, Joanna Obacz1,2, Florence Jouan1,2, Raphael Pineau1,2, Pierre Jean Le Reste1,2, Akram Obiedat6, Juhi Samal7, John Patterson8, Qingping Zheng9, Afshin Samali3, Abhay Pandit7, Boaz Tirosh6, Aristotelis Chatziioannou4,10, Eric Chevet1,2,*, Tony Avril1,2,*

1Proteostasis & Cancer Team, INSERM U1242 « Chemistry, Oncogenesis Stress Signaling », Université de Rennes, Rennes, France; 2Centre Eugène Marquis, Rennes, France; 3Apoptosis Research Centre, School of Natural Sciences, NUI Galway, Galway, Ireland; 4Institute of Biology, Medicinal Chemistry & Biotechnology, NHRF, Athens, Greece; 5Department of Biochemistry and Biotechnology, University of Thessaly, Larissa, Greece; 6Institute for Drug Research, School of Pharmacy, Faculty of Medicine, Hebrew University of Jerusalem, Jerusalem, Israel; 7CÚRAM, Centre for Research in Medical Devices, National University of Ireland, Galway, Ireland; 8Fosun Orinove PharmaTech Inc., 3537 Old Conejo Road, Suite 104, Newbury Park, CA 91320, USA. 9Fosun Orinove PharmaTech Inc., Suite 211, Building A4, 218 Xinghu St., Suzhou Industrial Park, Jiangsu 215123, China. 10e-NIOS PC, Kallithea-Athens, Greece.

Running title: IRE1 signaling and cancer cell stemness

Keywords: Cancer stem cells, Unfolded Protein Response, Endoplasmic reticulum

Correspondence to TA ([email protected]) or EC ([email protected])

144 Abstract

Cancer cell reprogramming (CCR) contributes to antineoplastic treatment resistance and disease recurrence through cancer stem cell re-emergence. Glioblastoma multiform (GBM) is a lethal, primary, central nervous system tumor, in which cancer cell reprogramming compounds prognostic severity. (referred to as IRE1 hereafter) is a major regulator of GBM development and is identified as an appealing therapeutic target. To validate IRE1 suitability as an antineoplastic pharmacological target, we tested its involvement in GBM cancer cell reprogramming. GBM patients’ transcriptomic analysis shows that high IRE1 activity correlates with the down-regulation of the main stemness transcription factors (TFs) SOX2, SALL2, POU3F2 and OLIG2. We pharmacologically and genetically recapitulate this effect in immortalized and primary GBM cell lines. We also report on a specific miRNA that acts through the IRE1-XBP1 signaling axis to produce these effects. Our results highlight a novel role of IRE1 as a negative regulator of CCR in GBM and highlights opportunities of informed IRE1 modulation utility in GBM therapy.

Statement of significance

Through a comprehensive bioinformatic, genetic and pharmacological study, we provide a detailed understanding of the role of IRE1 in GBM cell reprogramming uncovering a novel XBP1 dependent mechanism of differentiated GBM cell phenotype maintenance thus further documenting the pharmacological potential of targeting IRE1 in GBM.

145 Introduction

Glioblastoma multiforme (GBM) is the commonest primary solid tumor of the central nervous system with a dismal prognosis of median survival fewer than 2 years post diagnosis(1). GBMs are very heterogeneous differing in their resistance to chemotherapy and genetic instability(2), traits which impede therapeutic intervention, currently limited to maximal tumor resection followed by radiotherapy and chemotherapy with the alkylating agent temozolomide(3). GBM recurrence and therapeutic resistance can be attributed to cancer stem cell properties and differentiated-to-stem cell reprogramming capabilities that help GBM cells adapt to a hostile microenvironment(4). Genetic characteristics delineating tumor stem-like cells and differentiated tumor cells have been identified with nestin a prominent player alongside SOX2 in the former(5) and GFAP, VIM and YKL40 in the later(6). Interestingly nestin was shown to be overexpressed in GBM cells expressing a (hereafter called IRE1), a transducer of the Unfolded Protein Response (UPR)(7), which is an adaptive mechanism to high protein folding demand and hence misfolded protein accumulation that causes ER stress(8). IRE1 is a serine/ threonine kinase and endoribonuclease which upon ER stress oligomerizes and trans-auto phosphorylates mediating downstream signaling events that include JNK activation, the unconventional splicing of XBP1 mRNA to produce the transcription factor XBP1s and the degradation of a specific subset of miRNAs and mRNAs; a process called RNA regulated IRE1-dependent decay or RIDD(9). GBM cells are subjected to high metabolic demand, hypoxic stress, accelerated cell cycle events and GSC differentiation and reprogramming and they overcome these stresses partially through IRE1 signaling(10), which confers tumor cells with aggressive characteristics including i) the pro-tumoral remodeling of the tumor stroma with immune and endothelial cells and ii) high migration/invasion characteristics(11). Targeting the RNase activity of IRE1 negatively impacts tumor growth due to blocking of pro-survival cellular mechanisms mediated by XBP1 whilst IRE1 is involved in invasion, growth and vascularization as well as having dual and at times antagonistic functions of XBP1s and RIDD in GBM development(12,13). Since IRE1 and its target transcription factor XBP1s were shown to play major roles in cellular differentiation(14) and are involved in GBM development, we postulated that IRE1 activity may contribute GBM cell stemness regulation and thus we investigated the impact of IRE1 inhibition on the differentiation status of GBM cells provided that cancer cell

146 reprogramming contributes to antineoplastic treatment resistance and disease recurrence through cancer stem cells.

147 Results

IRE1 activity is associated with cancer differentiated state in GBM specimens We previously identified an IRE1 activity signature of 38 genes (IRE1_38; (11,13)) that was confronted to the TCGA GBM cohorts to stratify patients in two groups of high and low IRE1 activity; the former of which displaying a worse prognosis compared to the latter (11). Since we initially observed the expression of nestin in tumors deriving from U87 cells deficient for IRE1 signaling (7), we used the IRE1_38 signature to investigate the putative IRE1 dependent expression of markers of differentiation or stemness. We found that stem cell markers were upregulated whilst differentiation markers downregulated in tumors exhibiting low IRE1 activity as compared with those with high activity (Figure 1A, figure S1). This was representative in the case of stem markers BMI1, CD133 and nestin and differentiation markers SMA, vimentin and YKL40 (Figure 1B, figure S1). This led us to hypothesize that IRE1 could play a causal role in cancer stem cell biology in the maintenance of the differentiation phenotype. Subsequently we tested whether IRE1 activity could impact on the expression of transcription factors (TFs) controlling stemness. Indeed, we observed that stemness associated TFs were markedly decreased in tumors with high IRE1 activity (Figure 1C, figure S1), the four starkest of which were OLIG2, POU3F2, SALL2 and SOX2 (Figure 1D). To further document this phenomenon we established whether it could be recapitulated in multiple GBM lines. To this end we utilized common available U251 and U87 as well as primary RADH87 and RADH85(15) GBM lines genetically modified for IRE1. These were expressing a dominant negative (DN)(16) or truncated at residue Q780 (Q*)(11) form of IRE1, both lacking any ribonuclease activity. We observed a marked upregulation of OLIG2, POU3F2, SALL2 and SOX2 in the IRE1 signaling deficient lines compared to parental or WT IRE1 overexpressing cells (Figure 1E). As such we have demonstrated that in patients’ tumors high IRE1 activity correlates with a low level of reprogramming factors which is recapitulated in both commonly available and primary GBM cell lines.

Genetic perturbation of IRE1 effect on GBM cell reprogramming To further investigate the effects of IRE1 inhibition on GBM differentiated-to-stem reprogramming, we developed a culture system where cell lines normally grown in adherent 10% FCS-containing media (RADH primary and U87/U251 commonly available) were seeded in FCS-free neurosphere culture media and cell viability, cell number, phenotype as well as the ability of cells to form spheres from single cells (clonogenicity) were determined (Figure 2A,

148 figure S2A, C). Regarding the cellular phenotype, whilst no difference was seen between the parental cells and ones overexpressing a WT IRE1, DN or Q* IRE1 expressing cells readily formed spheres pertaining to a stem phenotype (Figure 2B, figure S2D). Furthermore, this observation was reinforce by the capacity of DN or Q* expressing cells to be grown as single cells over several passages (Figure 2C, figure S2B). Next, we evaluated the expression of OLIG2, POU3F2, SALL2 and SOX2 in cells grown in those conditions. Compared to parental, cells overexpressing a WT IRE1 showed no difference in the levels of any marker (Figure 2D). Contrastingly cells expressing either DN or Q* IRE1 showed uniform increase of stem and reprogramming markers whilst a decrease in differentiation markers (Figure 2D, figure S2E), supporting the hypothesis that IRE1 is essential for differentiated phenotype maintenance. This was further documented at the protein level were SOX2, nestin and A2B5 were all upregulated whilst NG2 downregulated in IRE1 deficient cells (Figure 2E, figure S2F). Moreover cells expressing DN or Q* IRE1 were significantly more clonogenic compared to parental or WT overexpressing lines (Figure 2F, figure S2G) compounding the stem phenotype in the absence of functional IRE1. We have thus demonstrated that in both commonly available and primary GBM cell lines, IRE1 genetic perturbation disturbs the differentiated cell phenotype, predisposing it to pre and post translational events that support a de- differentiation to a stem phenotype.

Pharmacological inhibition of IRE1 effect on GBM cell reprogramming As observed for the genetic inactivation of IRE1, the use of the salicylaldehyde IRE1 ribonuclease inhibitor MKC8866(17) phenocopied the effects of DN or Q* IRE1 on the capacity of cells to be grown as single cells over several passages (Figure 3A, figure S3A, B). When compared to the DMSO control, U251, RADH85, RADH87 cells treated with MKC8866 formed spheres more readily (Figure 3B, figure S3C, D), displayed higher mRNA levels of stem and reprogramming markers but lower mRNA levels of differentiation markers (Figure 3C, figure S3E) as well as higher protein levels of reprogramming markers compared to differentiation markers (Figure 3D, figure S3F). The clonogenic ability of the cells was also significantly upregulated in the presence of MKC8866 compared to control (Figure 3E, figure S3G). As such we show that both genetic and pharmacological perturbations of IRE1 lead to a loss of differentiated phenotype and genotype whilst at the same time pushes cells towards a sphere stem-like phenotype and genotype.

149 Role of IRE1 in GBM stem-to-differentiated state reprogramming Since IRE1 activity appears instrumental for the maintenance of the differentiated phenotype in GBM cells, we next tested whether IRE1 played a significant role in cancer stem cell differentiation. To this end, we utilized primary GBM lines grown in FCS-free neurosphere media(18) also modified for IRE1 with the Q* mutation(11), which were seeded in 10% FCS- containing media in the presence of the bone morphogenetic factor BMP4 to induce differentiation (Figure 4A, figure S4A). Genetic perturbation of IRE1 signaling did not result in a significant phenotype change in RNS lines when grown in FCS-containing media in the presence of BMP4 (Figure 4B, figure S4B). Moreover differentiation markers GFAP, O4 and TUJ1 showed slight decreases at the mRNA level in lines overexpressing WT or Q* forms of IRE1 (Figure 4C), an effect recapitulated at the protein level with further significance shown between the Q* and parental as opposed to WT and parental (Figure 4D). In the same line of observation, MKC8866 treatment yielded no phenotypic changes in RNS cells in the presence of 10% FCS media and BMP4 (Figure 4E) whilst differentiation markers although generally statistically significantly decreased at both the mRNA (Figure 4F) and protein level (Figure 4G) did not show the marked difference observed previously (Figures 2, 3). From this we can conclude that IRE1 signaling perturbation is modestly influencing cancer stem differentiation. However, it illuminates the hypothesis further that IRE1 is not the catalyst of GBM cell differentiation whilst it is of paramount importance in the maintenance of the differentiated phenotype.

XBP1s involvement in GBM cell reprogramming It is well described that IRE1 has a dual function mediated by its ribonuclease domain(8) through either XBP1 mRNA splicing or RIDD (19). We first correlated the expression of differentiation and stem markers in tumors stratified based on their XBP1s or RIDD status. We found that tumors exhibiting high XBP1s expressed low levels of stem markers and high levels of differentiation markers (Figure 5A). This was confirmed by investigating individual genes as well, where stem markers BMI1, CD133 and nestin showed no significant difference between high XBP1s and high RIDD activity tumors whilst differentiated tumor cell markers SMA, vimentin and YKL40 were significantly lower in RIDD high tumors as opposed to XBP1s high tumors (Figure 5B). This led us to pose that XBP1s rather than RIDD is the limiting factor in maintaining the differentiated GBM cell phenotype. Next, by monitoring the expression of 15 genes involved in reprogramming in XBP1s high or RIDD high tumors, we were able to discern

150 that reprogramming factors were downregulated in TCGA tumors with high XBP1s (Figure 5C). This was statistically confirmed when measuring the expression single genes OLIG2, POU3F2, SALL2 and SOX2 (Figure 5D). We then sought to evaluate whether XBP1s ablation was able to recapitulate this effect in commonly available (U251 and U87) and primary (RADH85 and RADH87) adherent differentiated GBM cell lines. XBP1s silencing led to the significant upregulation of almost all 4 TFs across all human lines (Figure 5F, figure S6A). We thus demonstrate that it is the IRE1-XBP1s signaling axis that controls the ability of GBM cells to maintain differentiation.

XBP1s-dependent expression of miR-148a prevents GBM cell reprogramming We next sought to identify the mediating factor between XBP1s induction and reprogramming TFs downregulation and hypothesized that these factors might be miRNAs (Figure 6A,(20)). Using miRNA sequencing, we first identified miRNA whose expression is dependent on the IRE1/XBP1s signaling axis and then investigated the expression of those miRNAs in tumors stratified based on high XBP1s or high RIDD (Figure 6B). This indicated that both miR-21 and miR-148a were the two miRNAs presenting the highest expression in XBP1s high tumors. Interestingly, miR-148a was identified previously as a transcriptional target of XBP1s using chromatin immunoprecipitation(21) and through sequence analysis we observed potential binding sites of miR148a on SOX2 (Figure S5). To further document the IRE1-dependent control of miR-148a expression, we first showed that miR-148a levels were decreased in IRE1 signaling deficient cells when comparing DN and Q* IRE1 expressing cells to parental (Figure 6D). Remarkably, this also corresponded to conditions where the upregulation of SOX2 and other TFs was observed (Figures 1-4). To consolidate the link between the expression of miR- 148a and XBP1s, we next measured miR-148a levels in cells silenced for XBP1s and we observed significant miR-148a downregulation (Figure 6E). To further validate this effect in vitro we artificially upregulated the expression of miR-148a in IRE1 deficient cells using miR- 148a mimics and observed a downregulation of SOX2 as well as the majority of the other TFs as opposed to the upregulation we would normally observe (Figure 6F, figure S6B). Furthermore to verify the link between XBP1s, miR-148a and SOX2 signaling we overexpressed XBP1s in IRE1 deficient and parental cells. In these conditions we observed the subsequent upregulation of miR-148a expression and the downregulation of SOX2 and other TFs in the majority of our human GBM cell lines (Figure 6G, figure S7A). A final validation of this signaling cascade was achieved by overexpressing XBP1s in IRE1 deficient cells and also treat the cells

151 with miR-148a inhibitors. As hypothesized, we observed an upregulation of XBP1s, a downregulation of miR-148a and the subsequent upregulation of SOX2. We have thus demonstrated that reprogramming TFs including SOX2 are downregulated by miR-148a, clarifying a relationship previously documented in the literature(22), and miR-148a is directly induced by XBP1s(21). Therefore we provide a direct signaling cascade that governs IRE1 influence on GBM cell reprogramming and their ability to de-differentiate in cancer stem-like cells.

IRE1-dependent control of GBM stemness reprogramming in vivo We next tested our working hypothesis, thus far validated in cell lines and in patients’ tumors, in a murine GBM model in order to obtain information in a living tumor progression setting. To this end we measured the levels of XBP1s, miR-148a and SOX2 in IRE1 knockout (KO) murine GL261 GBM cells. We found that the effect seen in human cells was recapitulated in murine GBM cells with miR-148a downregulation and subsequent SOX2 upregulation following an expected XBP1s downregulation (Figure 7A, figure S7B). We then proceeded to perform intracranial injections of GL261 parental or GL261 IRE1-KO cells in the brains of C57BL/6 mice. The tumor was allowed to grow and when reaching a critical mass for the mouse, the brain was harvested and multiple coronal sections were stained for the stem marker MSI1 (Figure 7B). Cells positive for stem markers were quantified to discern the effect of genetic IRE1 manipulation on tumor stemness in vivo (Figure 7C). This system was also probed pharmacologically using a similar approach. Intracranial injections of GL261 parental cells formed tumors which were allowed to grow for 14 days and then were surgically resected. A gel implant containing DMSO or MKC8866 was subsequently placed in the tumor cavity. The tumor was allowed to regrow and upon presentation of clinical symptoms such as rapid weight loss the mouse brain was removed and multiple coronal sections were stained for stem markers to analyze their expression by immunohistochemistry (Figure 7D). As is evident in the panels seen in Figures 7B,E brain sections through the tumor and the tumor periphery showed increased staining for stem marker MSI1 whilst not much difference was observed in the opposite hemisphere parenchyma. Multiple independent counts of different coronal planes for both genetic and pharmacological IRE1 inhibition quantified this effect showing a significant increase in stem marker staining in tumors treated with MKC8866 compared to tumors treated with DMSO (Figure 7E) as well as in tumors grown from GL261 IRE1-KO cells compared to those grown from GL261 parental cells (Figure 7C). We have thus

152 demonstrated that IRE1 inhibition leads to an increase in the presence of stem cell markers in an in vivo model of GBM.

153 Discussion

Our comprehensive analysis of the IRE1 signaling involvement in GBM cell reprogramming shows the IRE1-XBP1s-miR148a axis to be a major player in the maintenance of a differentiated cancer cell phenotype and paramount in absentia to the ability of non- terminally differentiated cells to revert to a stem phenotype. We thus expand on an IRE1 carcinogenesis repertoire that already includes angiogenesis and metastasis, and hence further assess its suitability as an antineoplastic pharmacological target. The role of IRE1 in posttranscriptional cell reprogramming towards adaptation in GBM is a the subject of a plethora of studies, with mRNA and miRNA stability being the focus of some investigations(10). In particular the increase of miR-17(23) in irradiated GBM stem cells as well as its potential anti-tumoral properties(11), the decrease of tumor suppressor miR-34a(24) in GBM and the increase of tumor growth promoting miR-96(25) have been documented. The analysis of the TCGA data has indeed shed additional light on a few miRNA candidates; miR-148a in particular, as especially interesting in GBM pathophysiology. Our focus on miR-148a as opposed to miR-21, another potential candidate our bioinformatics investigation uncovered, was guided by the demonstration that this miRNA was a transcriptional target of XBP1s(21) as well as our structural analysis of the sequence homology between miR148a and SOX2; one of the most important transcription factors in cell reprogramming (Figure S5). SOX2 and miR148a co-existed in studies examining neural crest stem cell physiology(26) and miR148a was reported as part of an investigation into the SOX2 miRNA response program in human stem cell lines(27). This of course does not exclude miR- 21 as a promising candidate in GBM pathophysiology. Rather, it offers further investigative scope to delve deeper in the functional network of miRNAs that may shape the response of IRE1 signaling in differentiation and cancer cell reprogramming. MiR-148a has been investigated in a few different malignancies with a meta-analysis showing it as a prognostic factor alongside miR-148b and miR-152(28,29). It has been shown that it promotes plasma cell differentiation by targeting germinal center TFs Mitf and Bach2(30), as well as to promote apoptosis in breast cancer cells(31). We here compound the importance of miR-148a in GBM and specifically its role in stem cell physiology and carcinogenesis in general alongside studies that have shown it to suppress the epithelial mesenchymal transition and metastasis of hepatoma cells(32), to increase glioma cell migration and invasion by downregulating GADD45A(33) and to be involved in the glabridin-induced inhibition of the cancer stem-like properties in hepatocellular carcinoma(34). We here offer an extensive investigation, and thus

154 valuable novel information, into the relationship that miR-148a has with XBP1s in the context of reprogramming as we show that miR-148a downregulates transcription factors involved in reprogramming of GBM cells upon activation by XBP1s. It is an established fact that heterogeneity is a major barrier to successful treatment in GBM. Moreover de-differentiation of GBM cells to a more stem-like phenotype compounds chemo-resistance. Both of these factors contribute to a basic truth of clinical trial design: it is of utmost importance to be able to stratify patients in order to maximize beneficial outcomes, as the status quo of taking broad clinical groups into consideration has caused many a promising therapeutic candidate to fail at phase 3 clinical trials or earlier(35). We have provided so far extensive research on the involvement of IRE1 in GBM carcinogenesis (11,12,16,36) and indeed patients’ stratification has been a founding driver in how we have conducted this investigation. We have previously shown that using our IRE1_38 signature, tumors may be stratified according to IRE1 activity (high IRE1 activity conferring worse prognosis) as well as according to the XBP1s or RIDD activity (XBP1s conferring worse prognosis)(11). Here we provide evidence that IRE1 inhibition and particularly XBP1s inhibition might promote differentiated tumor cells to revert to a more stem-like phenotype, an effect that has been linked with chemo-resistance and oncogenesis in general in GBM(37). As such our results point to two distinct opportunities for utilizing IRE1 signaling as therapeutic target in GBM. Firstly, patients with XBP1s low expressing tumors could benefit from IRE1 inhibition as GBM cells would have little capacity to utilize this signaling pathway for reverting to a stem phenotype and UPR disruption would overload a stress response mechanism that would already have to deal with a hostile microenvironment or with treatment. Indeed a recent study has shown that the IRE1-XBP1s axis is important in the adaptation to stress of leukaemic and healthy haematopoietic stem cells, as they enhance the ability of these cells to overcome ER stress and survive, promoting carcinogenesis(38). This information compounds the second and potentially more important outcome of our study that shows the attractiveness of IRE1 targeting therapeutics as adjuvant therapy alongside the currently established trident of surgery, chemotherapy and radiotherapy in GBM. The rationale would be that IRE1 targeting can sensitize GBM cells to therapy as it would weaken their responses and disallow them the time to adapt to the hostile conditions afforded by chemotherapy. As such patients would benefit by not only potential drops in the rates of tumor re-emergence but also of reduced need for repeated therapy doses which by default carry unfavorable toxicity profiles.

155 In conclusion, having demonstrated the importance of IRE1 in GBM development, prognosis and aggressiveness and having reviewed the role that GSCs play in similar characteristics we investigated the link between IRE1 signaling and GBM capacity in differentiation and reprogramming and sought to build upon the functionality of IRE1 in GBM pathophysiology. We described a direct link of IRE1 signaling to the reprogramming of differentiated cells back to stem cells in cellular models as well as correlating the two concepts in patient cohorts by demonstrating control of the transcription factors involved in reprogramming by IRE1 and XBP1s. Thereafter we delved further into the mechanisms of this IRE1 control and propose a novel concept where inhibition of XBP1s signaling induces the expression of transcription factors involved in GBM cell reprogramming and showcase the downstream miR148a signaling that makes this possible. Our work not only adds to the repertoire of IRE1 activity in GBM but also offers scope for patients’ stratification and combination therapy development with IRE1 targeting at its epicenter. It achieves this whilst reinforcing the need for strict pharmacovigilance when considering the risks of novel therapeutic design and builds upon our previous work in the field of translational neuro- oncology and ER biology to provide an ever more detailed landscape of IRE1 involvement and thus scrutinize the exploitability of the modulation of its function.

156 Materials & Methods

Reagents and antibodies All reagents not specified below were purchased from Sigma-Aldrich (St Quentin Fallavier, France). Recombinant human BMP-4 protein was obtained from R&D Systems (Lille, France); MKC8866 from Fosun Orinove; siRNA targeting XBP1 from Thermofisher Scientific; miR-148a inhibitors, mimics and controls (miRvana) were purchased from Thermo Fisher Scientific. For flow cytometry, antibodies against human GFAP, NG2, nestin, O4, SOX2, and TUJ1 were obtained from R&D Systems, Biotechne (Lille, France); anti-A2B5 from Miltenyi Biotec (Paris, France).

Cell culture and treatments U87MG (ATCC) and U251MG (Sigma, St Louis, MO, USA) cells were authenticated as recommended by AACR (http://aacrjournals.org/content/cell-line-authentication- information) and tested for the absence of mycoplasma using MycoAlert® (Lonza, Basel, Switzerland) or MycoFluor (Invitrogen, Carlsbad, CA, USA). Primary GBM cell lines were obtained as described in [AVRIL 2012]. U87, U251 and primary RADH GBM cells were grown in DMEM Glutamax (Invitrogen, Carlsbad, CA, USA) supplemented with 10% FBS. Primary GBM stem-like cell lines (RNS) were grown in DMEM/Ham’s:F12 (Life Technologies) supplemented with B27 and N2 additives (Life Technologies), EGF (20 ng/ml) and basic FGF (20 ng/ml) (Peprotech, Neuilly-sur Seine, France). Primary RADH85, RADH87, RNS85 and RNS87 were stably transfected at MOI = 0.3 with pCDH-CMV-MCS-EF1-Puro-copGFP (System biosciences) empty vector (EV), pCDH-CMV-MCS-EF1-Puro-copGFP containing IRE1alpha wild-type sequence (WT), or mutated sequence (Q780*). These puromycin, and polyclonal populations were tested for GFP expression. Transfections of GBM primary cell lines with IRE1 WT and Q780* were performed using Lipofectamine LTX (Thermo Fisher Scientific), according to the manufacturer's instructions. Murine GBM GL261 IRE1 KO was generated as described in [Akram 2019] and grown in DMEM Glutamax supplemented with 10% FBS.

Patients’ transcriptomic data from TCGA The publicly available GBM dataset of The Cancer Genome Atlas (TCGA) (Consortium et al., 2007; consortium, 2008) was assessed from the NCBI website platform (https://gdc- portal.nci.nih.gov) and was analyzed using the BioInfominer(39) analysis pipeline (E-Nios,

157 Greece, https://bioinfominer.com), which performs statistical and network analysis on various biological hierarchical vocabularies aiming to detect and rank significantly altered biological processes and the respective driver genes linking these processes. Genes were considered significantly differentially expressed if the p value was below 0.05. To analyze the miRNA database hierarchical clustering algorithms and Pearson correlation analyses were carried out using R packages.

Quantitative real-time PCR Total RNA was prepared using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Semi- quantitative analyses were carried out as previously described(12,16). All RNAs were reverse- transcribed with Maxima Reverse Transcriptase (Thermo Scientific, Waltham, MA, USA), according to manufacturer protocol. qPCR was performed with a StepOnePlus™ Real-Time PCR Systems from Applied Biosystems and the SYBR Green PCR Core reagents kit (TAKKARA). All RNAs for the miRNA investigation were transcribed using miScript RT kits and subsequent qPCR performed using miScript Primer Assays and miScript SYBR kits (QIAGEN). Experiments were performed with at least triplicates for each data point. Each sample was normalized on the basis of its expression of the actin gene. For quantitative PCR, the primer pairs used are described in Table S1.

Mouse Intracranial injections Eight-week old male C57BL/6 mice were housed in an animal care unit authorized by the French Ministries of Agriculture and Research (Biosit, Rennes, France - Agreement No. B35- 238-40). The protocol used was as previously described (7). Cell implantations were at 2 mm lateral to the bregma and 3 mm in depth using GL261 cells. Mice were daily clinically monitored and sacrificed twenty-four days post injection. Fourteen days post-injection the tumor formed was maximally removed without killing the animal and a plug infused with MKC8866 or DMSO control was implanted in the resection cavity. Fifteen days post plug implantation, or at first clinical signs, mouse brains were collected, fixed in 4% formaldehyde solution and paraffin embedded for histological analysis using anti-vimentin antibody (Interchim) to visualize the tumor masses.

158 Immunohistochemistry analyses Mouse tumor tissues were fixed in 10% neutral buffered formalin, embedded in paraffin, cut into 4- through a graded ethanol series. Endogenous peroxidase activity was quenched in 3% hydrogen peroxide (Roche) in PBS for 15 minutes. The IHC labeling were carried out using the H2P2 imaging platform of the faculty of Rennes. For cancer stem cell immunodetection, the sections were incubated with anti-MSI1 antibody (Merk Millipore), and analyzed with an Axioplan 2 epifluorescent microscope (Zeiss) equipped with a digital camera Axiocam (Zeiss).

Flow cytometry analyses Cells were washed in PBS 2% FBS and incubated with saturating concentrations of human immunoglobulins and fluorescent-labelled primary antibodies for 30 min at 4°C. Cells were then washed with PBS 2% FBS and analyzed by flow cytometry using a NovoCyte NovoSampler Pro (ACEA Biosciences). The population of interest was gated according to its FSC/SSC criteria. Data were analyzed with the NovoExpress software (ACEA Biosciences).

Statistical analyses Data are presented as mean ± SD or SEM (as indicated). Statistical significance (P < 0.05 or less) was determined using a paired or unpaired t-test or ANOVA as appropriate and performed using GraphPad Prism software (GraphPad Software, San Diego, CA, USA).

159 Acknowledgements

This work was funded by grants from Institut National du Cancer (INCa PLBIO), Fondation pour la Recherche Médicale (FRM, équipe labellisée 2018) to EC; EU H2020 MSCA ITN-675448 (TRAINERS) and MSCA RISE-734749 (INSPIRED) grants to AS, EC, EC; PHC Maimonide 2017- 2018 to BT and EC; the David R. Bloom Center for Pharmacy, the Dr. Adolph and Klara Brettler Center for Research in Pharmacology, German Israeli Fund (grant no. I-1471-414.13/2018) to BT; PROMISE, 12CHN 204 Bilateral Greece-China Research Program of the Hellenic General Secretariat of Research and Technology and the Chinese Ministry of Research and Technology sponsored by the Program “Competitiveness and Entrepreneurship,” Priority Health of the Peripheral Entrepreneurial Program of Attiki to AC. DD is a Marie Curie early stage researcher funded by EU H2020 MSCA ITN-675448 (TRAINERS). MMM was funded by the Irish Research Council and an ARED International PhD fellowship from “Région Bretagne”. JO was funded by a post-doctoral fellowship from “Région Bretagne”.

Conflict of interest

EC and AS are founders of Cell Stress Discoveries Ltd. AC is the founder of e-NIOS.

160 References 1. Ellis HP, Greenslade M, Powell B, Spiteri I, Sottoriva A, Kurian KM. Current Challenges in Glioblastoma: Intratumour Heterogeneity, Residual Disease, and Models to Predict Disease Recurrence. Front Oncol. Frontiers Media SA; 2015;5:251. 2. Shapiro JR, Yung W-KA, Shapiro WR. Isolation, Karyotype, and Clonal Growth of Heterogeneous Subpopulations of Human Malignant Gliomas. Cancer Res. 1981;41:2349 LP-2359. 3. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJBB, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. Massachusetts Medical Society ; 2005;352:987–96. 4. Hu B, Wang Q, Wang YA, Hua S, Sauvé C-EG, Ong D, et al. Epigenetic Activation of WNT5A Drives Glioblastoma Stem Cell Differentiation and Invasive Growth. Cell. 2016;167:1281–1295.e18. 5. Neradil J, Veselska R. Nestin as a marker of cancer stem cells. Cancer Sci. 2015/05/26. John Wiley & Sons, Ltd; 2015;106:803–11. 6. Widestrand Å, Faijerson J, Wilhelmsson U, Smith PLP, Li L, Sihlbom C, et al. Increased Neurogenesis and Astrogenesis from Neural Progenitor Cells Grafted in the

2007;25:2619–27. 7. Auf G, Jabouille A, Guérit S, Pineau R, Delugin M, Bouchecareilh M, et al. Inositol- requiring enzyme 1alpha is a key regulator of angiogenesis and invasion in malignant glioma. Proc Natl Acad Sci U S A. 2010/08/13. National Academy of Sciences; 2010;107:15553–8. 8. Doultsinos D, Avril T, Lhomond S, Dejeans N, Guédat P, Chevet E. Control of the Unfolded Protein Response in Health and Disease. SLAS Discov Adv Life Sci R&D. SAGE Publications Inc STM; 2017;22:2472555217701685. 9. Almanza A, Carlesso A, Chintha C, Creedican S, Doultsinos D, Leuzzi B, et al. Endoplasmic reticulum stress signalling – from basic mechanisms to clinical applications. FEBS J. Wiley/Blackwell (10.1111); 2018;0. 10. Obacz J, Avril T, Le Reste P-J, Urra H, Quillien V, Hetz C, et al. Endoplasmic reticulum proteostasis in glioblastoma—From molecular mechanisms to therapeutic perspectives. Sci Signal. 2017;10. 11. Lhomond S, Avril T, Dejeans N, Voutetakis K, Doultsinos D, McMahon M, et al. Dual

161 IRE1 RNase functions dictate glioblastoma development. EMBO Mol Med. Wiley- Blackwell; 2018;10:139–308. 12. Dejeans N, Pluquet O, Lhomond S, Grise F, Bouchecareilh M, Juin A, et al. Autocrine control of glioma cells adhesion and migration through IRE1alpha-mediated cleavage of SPARC mRNA. J Cell Sci. 2012/06/22. 2012;125:4278–87. 13. Pluquet O, Dejeans N, Bouchecareilh M, Lhomond S, Pineau R, Higa A, et al. Posttranscriptional regulation of PER1 underlies the oncogenic function of IREalpha. Cancer Res. 2013/06/12. 2013;73:4732–43. 14. Bettigole SE, Lis R, Adoro S, Lee AH, Spencer LA, Weller PF GL. The transcription factor XBP1 is selectively required for eosinophil differentiation. Nat Immunol. 2015;16:829– 37. 15. Avril T, Etcheverry A, Pineau R, Obacz J, Jegou G, Jouan F, et al. CD90 Expression Controls Migration and Predicts Dasatinib Response in Glioblastoma. Clin Cancer Res. 2017;23:7360 LP-7374. 16. Pluquet O, Dejeans N, Bouchecareilh M, Lhomond S, Pineau R, Higa A, et al. Posttranscriptional regulation of per1 underlies the oncogenic function of IRE?? Cancer Res. 2013;73:4732–43. 17. Volkmann K, Lucas JL, Vuga D, Wang X, Brumm D, Stiles C, et al. Potent and selective inhibitors of the inositol-requiring enzyme 1 endoribonuclease. J Biol Chem. 2011/02/10. American Society for Biochemistry and Molecular Biology; 2011;286:12743–55. 18. Avril T, Vauleon E, Hamlat A, Saikali S, Etcheverry A, Delmas C, et al. Human Glioblastoma Stem-Like Cells are More Sensitive to Allogeneic NK and T Cell-Mediated Killing Compared with Serum-Cultured Glioblastoma Cells. Brain Pathol. John Wiley & Sons, Ltd (10.1111); 2012;22:159–74. 19. Maurel M, Chevet E, Tavernier J, Gerlo S. Getting RIDD of RNA: IRE1 in cell fate regulation. Trends Biochem Sci. 2014/03/25. Elsevier Ltd; 2014;39:245–54. 20. McMahon M, Samali A, Chevet E. Regulation of the unfolded protein response by noncoding RNA. Am J Physiol Physiol. American Physiological Society; 2017;313:C243– 54. 21. Cho YM, Kim T-M, Hun Kim D, Hee Kim D, Jeong S-W, Kwon O-J. miR-148a is a downstream effector of X-box-binding protein 1 that silences Wnt10b during adipogenesis of 3T3-L1 cells. Exp &Amp; Mol Med. The Author(s); 2016;48:e226.

162 22. Lopez-Bertoni H, Lal B, Li A, Caplan M, Guerrero-Cázares H, Eberhart CG, et al. DNMT- dependent suppression of microRNA regulates the induction of GBM tumor- propagating phenotype by Oct4 and Sox2. Oncogene. 2014/10/20. 2015;34:3994– 4004. 23. Li H, Yang BB. Stress response of glioblastoma cells mediated by miR-17-5p targeting PTEN and the passenger strand miR-17-3p targeting MDM2. Oncotarget. 2013/02/09. Impact Journals LLC; 2013;3:1653–68. 24. Li Y, Guessous F, Zhang Y, Dipierro C, Kefas B, Johnson E, et al. MicroRNA-34a inhibits glioblastoma growth by targeting multiple oncogenes. Cancer Res. 2009/09/22. 2009;69:7569–76. 25. Yan Z, Wang J, Wang C, Jiao Y, Qi W, Che S. miR- -catenin regulatory circuitry promotes glioma growth. FEBS Lett. John Wiley & Sons, Ltd; 2014;588:3038– 46. 26. Ichi S, Costa FF, Bischof JM, Nakazaki H, Shen Y-W, Boshnjaku V, et al. Folic acid remodels chromatin on Hes1 and Neurog2 promoters during caudal neural tube development. J Biol Chem. American Society for Biochemistry and Molecular Biology; 2010;285:36922–32. 27. Vencken SF, Sethupathy P, Blackshields G, Spillane C, Elbaruni S, Sheils O, et al. An integrated analysis of the SOX2 microRNA response program in human pluripotent and nullipotent stem cell lines. BMC Genomics. BioMed Central; 2014;15:711. 28. Miao C, Zhang J, Zhao K, Liang C, Xu A, Zhu J, et al. The significance of microRNA- 148/152 family as a prognostic factor in multiple human malignancies: a meta- analysis. Oncotarget. Impact Journals LLC; 2017;8:43344–55. 29. Friedrich M, Pracht K, Mashreghi M-F, Jäck H-M, Radbruch A, Seliger B. The role of the miR-148/-152 family in physiology and disease. Eur J Immunol. Wiley-Blackwell; 2017;47:2026–38. 30. Porstner M, Winkelmann R, Daum P, Schmid J, Pracht K, Côrte-Real J, et al. miR-148a promotes plasma cell differentiation and targets the germinal center transcription factors Mitf and Bach2. Eur J Immunol. Wiley-Blackwell; 2015;45:1206–15. 31. Li Q, Ren P, Shi P, Chen Y, Xiang F, Zhang L, et al. MicroRNA-148a promotes apoptosis and suppresses growth of breast cancer cells by targeting B-cell lymphoma 2. Anticancer Drugs. 2017;28. 32. Zhang J-P, Zeng C, Xu L, Gong J, Fang J-H, Zhuang S-M. MicroRNA-148a suppresses the

163 epithelial–mesenchymal transition and metastasis of hepatoma cells by targeting Met/Snail signaling. Oncogene. Macmillan Publishers Limited; 2013;33:4069. 33. Cui D, Sajan P, Shi J, Shen Y, Wang K, Deng X, et al. MiR-148a increases glioma cell migration and invasion by downregulating GADD45A in human gliomas with IDH1 R132H mutations. Oncotarget. Impact Journals LLC; 2017;8:25345–61. 34. Jiang F, Mu J, Wang X, Ye X, Si L, Ning S, et al. The Repressive Effect of miR-148a on TGF beta-SMADs Signal Pathway Is Involved in the Glabridin-Induced Inhibition of the Cancer Stem Cells-Like Properties in Hepatocellular Carcinoma Cells. Yang L-Y, editor. PLoS One. Public Library of Science; 2014;9:e96698. 35. von Neubeck C, Seidlitz A, Kitzler HH, Beuthien-Baumann B, Krause M. Glioblastoma multiforme: emerging treatments and stratification markers beyond new drugs. Br J Radiol. 2015/08/03. The British Institute of Radiology.; 2015;88:20150354. 36. Lhomond S, Pallares N, Barroso K, Schmit K, Dejeans N, Fazli H, et al. Adaptation of the secretory pathway in cancer through IRE1 signaling. Methods Mol Biol. 2015/03/26. 2015;1292:177–94. 37. Gargiulo G, Cesaroni M, Serresi M, de Vries N, Hulsman D, Bruggeman SW, et al. In Vivo RNAi Screen for BMI1 Targets Identifies TGF- -ER Stress Pathways as Key Regulators of Neural- and Malignant Glioma-Stem Cell Homeostasis. Cancer Cell. Cell Press; 2013;23:660–76. 38. Liu L, Zhao M, Jin X, Ney G, Yang KB, Peng F, et al. Adaptive endoplasmic reticulum –XBP1 preserves self-renewal of haematopoietic and pre- leukaemic stem cells. Nat Cell Biol. Nature Publishing Group; 2019;21:328–37. 39. Georgiadis P, Liampa I, Hebels DG, Krauskopf J, Chatziioannou A, Valavanis I, et al. Evolving DNA methylation and gene expression markers of B-cell chronic lymphocytic leukemia are present in pre-diagnostic blood samples more than 10 years prior to diagnosis. BMC Genomics. BioMed Central; 2017;18:728.

164 Figure Legends Figure 1. IRE1 activity is associated with cancer differentiated state in GBM specimens. A) Hierarchical clustering of GBM patients (TCGA cohort) based on high or low IRE1 activity confronted to differentiation and stem signatures derived from literature. B) mRNA expression of BMI1, CD133, nestin stem cell markers and SMA, vimentin, YKL40 differentiated cell markers based on microarray fluorescence intensity in high and low IRE1 activity tumors. C) Hierarchical clustering of GBM patients (TCGA cohort) based on high or low IRE1 activity confronted to a reprogramming gene signature derived from literature. D) mRNA levels of reprogramming transcription factors OLIG2, POU3F2, SALL2, and SOX2 based on microarray fluorescence intensity in high and low IRE1 activity tumors. E) mRNA levels of reprogramming TFs in U251, RADH85 and RADH87 lines expressing WT, DN or Q* forms of IRE1 normalized to parental.

Figure 2. Genetic perturbation of IRE1 effect on GBM cell reprogramming. A) Schematic representation of GBM cell working model of differentiated to stem cell phenotype culture. B) Phenotypic characterization of U251/ RADH85/ RADH87 parental and overexpressing WT, DN or Q* forms of IRE1 when grown in neurosphere media. C) Differentiated GBM cell lines U251, RADH85 and RADH87 were cultured in neurosphere medium and were passaged every 14 days. If the number of cells was under the initial number of cells seeded (106), the culture was stopped. D) Heat map representation of fold change of mRNA expression of genes involved in reprogramming, stemness and differentiation normalized to parental in U251, RADH85, RADH87 lines expressing WT, DN or Q* forms of IRE1 when grown in neurosphere media. E) Protein expression of reprogramming, stemness and differentiation markers in these lines normalized to parental. F) Clonogenicity of differentiated lines expressing WT, DN or Q* forms of IRE1 normalized to parental when grown in neurosphere media.

Figure 3. Pharmacological inhibition of IRE1 effect on GBM cell reprogramming. A) Differentiated GBM cell lines U251, RADH85 and RADH87 were cultured in neurosphere medium in the presence of MKC8866, and were passaged every 14 days. If the number of cells was under the initial number of cells seeded (106), the culture is stopped. B) Phenotypic characterization of parental adherent (RADH85/87 and U251) lines through culture in neurosphere medium treated with MKC8866 or DMSO. C) Heat map representation of fold change of mRNA expression of genes involved in reprogramming, stemness and differentiation normalized to parental in U251, RADH85, and RADH87 lines when grown in neurosphere

165 media in the presence of MKC8866 or DMSO. D) Protein expression of reprogramming, stemness and differentiation markers in these lines normalized to parental. E) Quantification of clonogenicity of single cell parental, WT or Q* IRE1 expressing RADH85/87 and U251 lines when seeded in serum-free medium in the presence of MKC8866 or DMSO.

Figure 4. Role of IRE1 in GBM stem-to-differentiated state reprogramming. A) Schematic representation of GBM cell working model of stem-to-differentiated cell phenotype culture. B) Phenotypic characterization of RNS85/87 parental and overexpressing WT or Q* forms of IRE1 when grown in FCS containing media. C) Heat map representation of fold change of mRNA expression of genes involved in differentiation normalized to parental in RNS85, RNS87 lines expressing WT or Q* forms of IRE1 when grown in FCS containing media. D) Protein expression of differentiation markers in these lines normalized to parental. E) Phenotypic characterization of parental RNS85/87 lines through culture in FCS-containing medium treated with MKC8866 or DMSO. F) Heat map representation of fold change of mRNA expression of genes involved in differentiation in parental RNS85, RNS87 lines when grown in FCS containing media in the presence of MKC8866 or DMSO. G) Protein expression of differentiation markers in these lines in the presence of MKC8866 or DMSO.

Figure 5. XBP1s involvement in GBM cell reprogramming. A) Hierarchical clustering of genes involved in differentiation and stemness in GBM (TCGA cohort) based on high XBP1s or high RIDD activity (blue low levels, red high levels). B) mRNA expression of BMI1, CD133, nestin stem cell markers and SMA, vimentin, YKL40 differentiated cell markers based on microarray fluorescence intensity in high XBP1s and high RIDD activity tumors. C) Hierarchical clustering of genes involved in reprogramming in GBM (TCGA cohort) based on high XBP1s or high RIDD activity (blue low levels, red high levels). D) mRNA expression of SOX2, POU3F2, OLIG2, SALL2 reprogramming TFs based on microarray fluorescence intensity in high XBP1s and high RIDD activity tumors. E) mRNA levels of SOX2 upon XBP1s silencing in primary and classical adherent GBM lines (RADH85/87 and U251 respectively).

Figure 6. XBP1s-dependent expression of miR-148a prevents GBM cell reprogramming. A) Schematic representation of hypothesis of effect of IRE1 signaling on reprogramming TFs. B) Hierarchical clustering of miRNAs in GBM (TCGA cohort) confronted to high XBP1s or high RIDD activity (blue low levels, red high levels) with the best 5 candidates shown. C) mRNA expression

166 of miR148a based on microarray fluorescence intensity in high XBP1s and high RIDD activity tumors. D) miR148a expression in adherent lines U251, RADH85/87 expressing DN or Q* forms of IRE1 normalized to parental. E) miR148a expression in adherent lines U251, RADH85/87 transiently deficient for XBP1s through siRNA transfection compared to control. F) SOX2 and miR148a expression levels in RADH85 IRE1 Q* expressing cells in the presence of miR148a mimic compared to control. G) XBP1s, SOX2 and miR148a expression levels in RADH85 IRE1 Q* expressing cells, over-expressing XBP1s compared to control. H) XBP1s, SOX2 and miR148a expression levels in RADH85 IRE1 Q* expressing cells, over-expressing XBP1s, in the presence of miR148a compared to control.

Figure 7. IRE1-dependent control of GBM stemness reprogramming in vivo. A) XBP1s, SOX2 and miR148a expression levels in GL261 IRE1 KO cells, compared to parental. B) Representative sections of tumors grown from GL261 parental or GL261 IRE1-KO cells injected in the brain of orthotopic syngeneic mouse model, stained for MSI. Sections from tumor body, tumor periphery and opposite to tumor brain parenchyma shown. C) Quantification of MSI1 positive cells (3-4 tumors/condition, >40 fields/tumor/condition quantified, three independent counts). D) Orthotopic syngeneic mouse GBM model and peri-operative treatment with MKC8866 (plug). E) MSI1 staining (stem marker) of tumor body or periphery and opposite to tumor site hemisphere parenchyma. This was performed in GL261 derived tumors treated with control or MKC8866 plugs. F) Quantification of MSI1 positive cells (3 tumors/condition, >40 fields/tumor/condition quantified, 3 independent counts).

167

168

169

170

171

172 173

174 Supplemental material

Figure S1. Exploratory Data Analysis (EDA) of IRE1 signature association with the mRNA expression level of differentiation- and stem-related genes in TCGA-GBM patient cohort. Normalized Affymetrix U133 microarray data (527 cases) and scaled by gene log2CPM RNAseq data (156 cases) were exploited for gene expression profiling. (A) Unsupervised hierarchical clustering analysis (pearson complete linkage on standardized microarray data) representing the association of 38-IRE1 signature with the expression profile of differentiation and stem markers. The XBP1s component of IRE1 signature is represented by the red bar and the RIDD component by the blue one. (B) Contingency tables of the distribution of IRE1activity groups (groups in rows) by the distribution of differentiation- and stem-activity groups (groups in columns), respectively. Patient grouping was based on the quantile-oriented XBP1s score (IRE1_sign score) (Lhomond et. al., 2018) and the differentiation/stem enrichment score (Diff/Stem_sign score). The latter denotes the ratio between the number of over- (or down-expressed) differentiation or stem markers to the total number of differentiation or stem markers were examined, respectively. Each differentiation or stem marker was appointed as over- or down-expressed when its gene expression in the sample was larger than the median gene expression in cohort. Mid- state represents equal distribution between over- and down-expressed markers. Chi-square hypothesis tests of independence were performed to analyze the correlation between IRE1activity and differentiation and stem marker activity groups. The p-value of each statistical test is surrounded by a red box. (C) Hierarchical clustering of TCGA-GBM patients (RNAseq data) based on their IRE1_sign score coupled with the differentiation/stem enrichment score. Samples were clustered using Euclidean distance metrics with Ward.D2 linkage. (D) Hierarchical clustering of TCGA-GBM patients (RNAseq data) using the expression profile of IRE1 signature coupled with the expression levels of differentiation and stem markers. Complete linkage and Pearson correlation was used as distance measure. Two crosstabs are also used to aggregate and jointly display the distribution of IRE1activity groups with the distribution of differentiation- and stem-activity groups. Fisher's Exact Test was additionally performed because of the small number of cases. IRE1 high activity is statistically significant correlated with high differentiation activity and low stem activity and vice versa. p-value <0.0001 and p-value <0.001 is denoted as .000 and .001, respectively, by the SPSS statistical software.

Figure S2. Genetic perturbation of IRE1 effect on GBM cell reprogramming in U87 cells. A) mRNA expression levels of reprogramming TFs in U87 cells expressing a dominant negative form of IRE1 normalised to parental. B) Differentiated GBM cell line U87 was cultured in neurosphere medium and passaged every 14 days. If the number of cells was under the initial number of cells seeded (106), the culture was stopped. C) Percentage of living cells and cell number per passage in differentiated U87/251 and RADH85/87 GBM cells when grown in neurosphere media and carrying a WT, DN or Q* form of IRE1 compared to parental. D) Phenotypic characterization of U87 parental and overexpressing DN form of IRE1 when grown in neurosphere media. E) Heat map representation of fold change of mRNA expression of genes involved in reprogramming, stemness and differentiation normalized to parental in U87 lines expressing DN form of IRE1 when grown in neurosphere media. F) Protein expression of reprogramming, stemness and differentiation markers in these lines normalized to parental. G) Clonogenicity of differentiated lines expressing DN form of IRE1 normalized to parental when grown in neurosphere media.

175 Figure S3. Pharmacological inhibition of IRE1 effect on GBM cell reprogramming in U87 cells. A) Schematic representation of GBM cell working model of differentiated to stem cell phenotype culture in the presence of MKC8866. B) Differentiated GBM cell line U87 was cultured in neurosphere medium in the presence of MKC8866, and passaged every 14 days. If the number of cells was under the initial number of cells seeded (106), the culture is stopped. C) Percentage of living cells and cell number per passage in differentiated U87/251 and RADH85/87 GBM cells when grown in neurosphere media in the presence of MKC8866. D) Phenotypic characterization of parental adherent U87 lines through culture in neurosphere medium treated with MKC8866 or DMSO. E) Heat map representation of fold change of mRNA expression of genes involved in reprogramming, stemness and differentiation normalized to parental in U87 lines when grown in neurosphere media in the presence of MKC8866 or DMSO. D) Protein expression of reprogramming, stemness and differentiation markers in these lines normalized to parental. E) Quantification of clonogenicity of single cell parental U87 lines when seeded in serum-free medium in the presence of MKC8866 or DMSO.

Figure S4. Role of IRE1 in GBM stem-to-differentiated state reprogramming proof of concept. A) Schematic representation of GBM cell working model of stem-to-differentiated cell phenotype culture in the presence or absence of MKC8866. B) Phenotypic characterization of RNS85/87 parental and overexpressing WT or Q* forms of IRE1 when grown in FCS containing media (RNS + BMP4) with comparator panels when grown in neurosphere media (RNS). C) Levels of BMP receptor in RNS85/87 normalized to RNS85 to predict responsiveness to BMP4. D) Protein expression of differentiation markers in RNS85/87 over 10 days in the presence of BMP4. E) Protein expression of differentiation markers in RNS85/87 in the presence of BMP4 for the final experimental conditions as dictated by D). F) Heat map representation of fold change of mRNA expression of genes in E).

Figure S5. Sequence based evidence of potential SOX2/miR148a interaction. A) miR148a secondary structure. B) miR148a 5’ and 3’ strand sequences. C) SOX2 mRNA map with indications of sequence homology and hence binding site probability of both 3’ and 5’ strands of miR148a. Two sites are identified per strand, with miR148a 5’ potentially binding on both the 5’ and 3’ UTR of SOX2 mRNA whilst miR148a 3’ occupying two sites on the 3’ UTR. D) Species homology across human, monkey, rat and mouse of SOX2 and miR148a sequence complementarity of all four sites identified in C).

Figure S6. XBP1s downregulation and miR-148a overexpression have opposite effects on reprogramming TFs in GBM cells. A) SALL2, POU3F2, OLIG2 mRNA expression in parental adherent lines U251, RADH85/87 transiently deficient for XBP1s through siRNA transfection compared to control. B) SALL2, POU3F2, OLIG2, SOX2 mRNA expression levels in U251, RADH85/87 IRE1 DN and Q* expressing cells in the presence of miR148a mimic compared to control.

Figure S7. XBP1s controls expression levels of TFs involved in reprogramming in multiple human and murine GBM cell lines. A) SOX2, SALL2, POU3F2 and OLIG2 mRNA expression levels in U87/251 and RADH85/87 parental or IRE1 DN/Q* expressing cells, over-expressing XBP1s compared to control. B) OLIG2, POU3F2, SALL2 mRNA expression levels in GL261 cells, null for IRE1 compared to parental.

176 177 178 179 180 181 182 183 Chapter 5: Conclusions and Contributions

In this fifth chapter we have established that the IRE1-XBP1-miR148a axis is a major regulator of transcription factors in general and SOX2 in particular, involved in GBM cell reprogramming. We have combined comprehensive bioinformatic, genetic and pharmacological approaches to test the significance of IRE1 downregulation in GBM cell systems. Having added to the repertoire of IRE1 in GBM carcinogenesis we have observed findings that offer scope for patient stratification and combination therapy development, whilst providing reason for careful novel drug discovery and IRE1 modulation in GBM.

The author of this thesis contributed to the publication of “IRE1 signalling maintains glioblastoma cell differentiation through the XBP1s/miR-148a mediated repression of stemness transcription factors” the following: Original text for the entire manuscript including materials and methods, figure legends and supplemental information11. Experimental work amounting to data presented in figures 2-712 and equivalent supplemental figures as well as figure assembly especially but not exclusively figures 4, 6.

Particular thanks to Dr T. Avril for his substantial contribution in conception and realisation of this body of work in its every facet including experimentation, figure construction and data supplementation.

11 Special thanks to Dr T. Avril and Dr E. Chevet for their help in shaping the manuscript to take its final form. 12 Special thanks to Dr j. Obacz for carrying out some quantitative PCR completing figure 2, to M. McMahon for producing the list of miRNA candidates investigated, to A. Obiedat and Prof B. Tirosh for their donation of IRE1 KO GL261 cells and to F. Jouan and R. Pineau for their work with the mouse model.

184 Chapter 6 Novel IRE1 inhibitors chemo-sensitise Glioblastoma cells to Temozolomide

185 Chapter 6: Foreword

In chapter 5, we explored the role of IRE1 in the maintenance of the GBM differentiated phenotype. Through this investigation we uncovered a regulatory pathway of transcription factors involved in GBM cell reprogramming through XBP1s signalling and miRNA148a induction. These results reinforced IRE1’s position as an interesting pharmacological target in GBM with a particular focus in sensitising GBM cells to current treatments. In chapter 6, we shall attempt to do just this and design novel IRE1 modulators that sensitise GBM cells to TMZ. To this end we will use a novel approach where fragments of IRE1 itself will be used as structural and binding templates to construct compounds that could bind the IRE1 kinase domain in silico. Since these are based on the IRE1 amino acid sequence they could potentially bear specificity due to sequence and structural homology with the binding pocket of IRE1. We shall observe how a final library of compounds consisting of small peptides, FDA approved and non FDA approved molecules was generated and then tested in vitro and in cell systems. We will follow a logical sequence of events from library formation to translation to a GBM model, starting from establishing whether these compounds affected IRE1 activity in vitro in a clean experimental cleavage assay. Thereafter, having established that these compounds affect IRE1 RNase activity in vitro we shall query the effect of the kinase binding compounds on IRE1 phosphorylation and their effect on XBP1 splicing in the presence of two ER stressors in both commonly available and primary GBM cell lines. Upon showing that they do indeed affect IRE1 activity in cells we will define their ability to sensitise GBM cells to TMZ using toxicity profiles and escalation doses in both primary and commonly available lines, thus combining cell models described in chapter 4 and the concept of chemosensitivity proposed in chapter 5. As such through computational, chemical and biological means, we will attempt to fulfil the third and final major objective of this thesis.

186 Novel IRE1 inhibitors chemo-sensitise Glioblastoma cells to Temozolomide

Dimitrios Doultsinos1, 2, Antonio Carlesso3, James C. Paton4, Adrienne W. Paton4, Leif A. Eriksson3,*, Eric Chevet1, 2,*

1Proteostasis & Cancer Team INSERM U1242 « Chemistry, Oncogenesis Stress Signaling », Université de Rennes, Rennes, France. 2Centre de Lutte contre le Cancer Eugène Marquis, Rennes, France. 3Department of Chemistry and Molecular Biology, University of Gothenburg, Göteborg, Sweden. 4Research Centre for Infectious Diseases, Department of Molecular and Biomedical Science, University of Adelaide, Adelaide, 5005, Australia.

*Equally contributed and corresponding authors EC [email protected] and LAE [email protected]

Keywords: Endoplasmic Reticulum • IRE1 • Inhibitors • glioblastoma • Unfolded Protein Response

Running title: New IRE1 inhibitors in GBM

Financial support: This work was supported by grants from Institut National du Cancer (INCa; PLBIO), Agence Nationale de la Recherche (ANR; ERAAT) and Fondation pour la Recherche Médicale (FRM; Equipe labellisée 2018) to EC. DD, CC, AS, LAE and EC are funded by EU H2020 MSCA ITN-675448 (TRAINERS) and MSCA RISE-734749 (INSPIRED) grants. The Swedish Research Council (VR) and the Swedish National Infrastructure for Computing (SNIC) are gratefully acknowledged for funding and allocations of computing time, respectively (LAE).

Conflict of Interest: LEA and EC are founders of Cell Stress Discoveries Ltd.

187 Abstract Inositol Requiring Enzyme 1 (IRE1) is a bifunctional serine/threonine kinase and endoribonuclease that is a major mediator of the Unfolded Protein Response (UPR) during endoplasmic reticulum (ER) stress. Tumour cells experience ER stress due to adverse environmental cues such as hypoxia or nutrient shortage and high metabolic/protein folding demand. To cope with those stresses, cancer cells utilise IRE1 signalling as an adaptive mechanism. Here we report the discovery of both FDA and non-FDA approved compounds as IRE1 inhibitors identified through a structural exploration of the IRE1 kinase domain. These inhibitors were tested for their ability to sensitise glioblastoma (GBM) cells to chemotherapy. We show that all molecules identified sensitise glioblastoma cells to the standard of care chemotherapy temozolomide (TMZ). These results support the attractiveness of IRE1 as an adjuvant therapeutic target in GBM; a common and wholly fatal diagnosis. In addition, they provide scope for quickly testing IRE1 inhibitor suitability in GBM as adjuvant therapy due to their current status for clinical use.

188 Introduction

GBM is the most common primary CNS tumour, displaying high levels of aggressiveness, recurrence and heterogeneity; traits that contribute to a dismal prognosis of an average of 1.5 year survival post diagnosis. The standard of care comprises maximal safe resection of the tumour followed by a combination of irradiation and chemotherapy with the alkylating agent temozolomide; however, all patients succumb to the disease(1). GBM cells, as with most solid tumours, survives in a hostile environment which includes hypoxia, nutrient shortage, necrosis and immune infiltration, as well as having to cope with a high metabolic turnover and protein synthesis demand(2). As such, the Unfolded Protein Response (UPR) is inextricably linked to GBM pathophysiology(3). It has been shown that IRE1, a major UPR transducer, in particular plays a decisive role in tumorigenesis and aggressiveness as through XBP1s signalling it is promoting tumour infiltration by immune cells, angiogenesis and invasion. GBM tumours displaying high levels of IRE1/XBP1 activity have a worse prognosis than those with low activity(4). This pertains to the possibility that attenuating IRE1 activity could lead to sensitization of tumours to current therapies as GBM cells would exhibit reduced capacity to cope with the hostile environment (Scheme 1). Indeed such studies have been performed in Triple Negative Breast Cancer (TNBC) showing that inhibition of IRE1 RNase activity with salicylaldehyde MKC8866 increased paclitaxel-dependent attenuation of TNBC development in mouse xenograft models(5). Further to this, MKC8866 treatment greatly enhanced the efficacy of docetaxel in regressing MYC- overexpressing tumours in breast cancer PDX models(6). This inhibitor is currently tested on other types of cancers.

IRE1 activity inhibition can be mediated by compounds targeting either the ATP-binding kinase domain or directly with the RNase domain. Direct RNase pharmacological inhibitors include 4μ8c, STF-083010, toyocamycin and a series of MKC compounds, whilst kinase pharmacological inhibitors that in turn inhibit the RNase include compound 3 and, although unclear as to its effect on IRE1 activity, sunitinib(7). The description of an allosteric IRE1 RNase inhibitory mechanism by ATP competitive ligands was provided through the discovery of Kinase inhibiting RNase attenuators (KIRAs) showing that inhibition of the kinase site may have an

189 inhibitory effect on the RNase activity(8,9). In addition to those pharmacological inhibitors, research from our laboratory discovered that peptide fragments derived from IRE1 cytosolic domain itself, affected its oligomerisation and subsequent RNase activity(10,11). Here, using a structural homology approach, we identify compounds that resemble IRE1 peptide fragments and demonstrate that they have an impact on IRE1 activity in vitro and in human cell models of GBM, whilst at the same time addressing the clinical relevance of IRE1 inhibition in GBM by demonstrating that administration of these compounds sensitise human of GBM cells to temozolomide (TMZ).

190 Materials and Methods Materials – Recombinant human IRE1 cytosolic domain was from Sinobiological (11905-HNCB). The fluorescent probe used for the in vitro IRE1 RNase assay was from Eurogentec (Cy3-CAUGUCCGCAGCGCAUG -BHQ3). Tunicamycin was purchased from Calbiochem (Merck KGaA, Darmstadt, Germany). SubAB toxin was supplied by Professors J. C. and A. W. Paton (University of Adelaide). All the other inhibitors were synthesised by and purchased from companies outlined as follows. Peptides, F6P1 and F6P2 were synthesized and purchased from Biomatik (Cambridge, Canada). Methotrexate, folinic acid and cefoperazone were from Apollo Scientific (Stockport, UK), Fludarabine Phosphate was from Selleck Chemicals (Munich, Germany), Z4 and Z6 were synthesised and purchased from Enamine (Riga, Latvia).

IRE1-mediated in vitro RNase assay - Peptides were diluted in Reaction buffer, organic molecules in minimal volume of DMSO and subsequently re-diluted in reaction buffer. Maximum volume of DMSO per reaction never exceeded 1%. Reaction volume was 25μl. Recombinant IRE1 was incubated at room temperature for 10 minutes with varying concentrations (0-50 μM) of inhibitor and reaction buffer. Subsequently equal volume of mixture of reaction buffer, ATP, DTT and fluorescent probe were added to each sample and fluorescence was read in a 96 well, black, flat bottom plate every minute for 25 minutes, at 37°C, in a Tecan 200 plate reader.

Cell culture and treatments - U87MG (ATCC) cells were authenticated as recommended by AACR (http://aacrjournals.org/content/cell-line-authentication- information) and tested for the absence of mycoplasma using MycoAlert® (Lonza, Basel, Switzerland) or MycoFluor (Invitrogen, Carlsbad, CA, USA). U87MG were grown in DMEM Glutamax (Invitrogen, Carlsbad, CA, USA) supplemented with 10% FBS. Primary GBM RADH87 cells were generated as previously described(12). For XBP1s induction experiments, tunicamycin was used at 1 μg/ml for the indicated periods of time. SubAB toxin was used at 1μg/ml up to 1.5 hours. Inhibitors investigated were used at concentrations indicated in figure S14. For inhibitor cell toxicity assays cells were plated in 96 well plates at 5000 cells per well and treated with 0, 5, 10, 25, 50, 100, 250, 500, 1000 and 2500 μM concentrations of each inhibitor. After 6 days of incubation WST1 reagent (Roche) was added to each well

191 and post 4 hour incubation the plate was read using a Tecan 200 colorimeter. For TMZ sensitivity assays cells were plated in a 96 well plate at 5000 cells per well and co-treated with 0, 5, 10, 25, 50, 100, 250, 500, 1000 and 2500 μM of TMZ plus a non- toxic dose of inhibitor described in figure S14. After 6 days of incubation WST1 reagent (Roche) was added to each well and post 4 hour incubation the plate was read using a Tecan 200 colorimeter. Peptides F6P1 and F6P2 were transfected in HEK293T cells using Pro-JectTM Protein Transfection Reagent Kit (Thermofisher Scientific, USA) according to the manufacturer’s instructions.

Statistical analyses - Data are presented as mean ± SD. Statistical significance (P < 0.05 or less) was determined using unpaired tǦtests or ANOVA as appropriate and performed using GraphPad Prism software (GraphPad Software, San Diego, CA, USA). Toxicity and sensitisation curve extrapolation was performed using curve fit hypotheses by GraphPad Prism software (GraphPad Software, San Diego, CA, USA).

In silico docking - Libraries of compounds were formed using Maestro Suites 2015- 2018 (Schrödinger Release 2018-4: Schrödinger, LLC, New York, NY, 2018) and specifically utilising tools within this programme such as LigPrep, Glide Dock, Pharmacophore Hypothesis Generation, Glide grid generation and Protein preparation. Structural exploration and peptide synthesis were carried out using software Supermimic(13) and MOE (MOE 2018.01: Chemical computing group, Montreal Canada), respectively. UCSF Chimera(14) was used for image generation developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311. Hypotheses were tested against ZINC and FDA databases using the Supercomputer cluster Hebbe at the C3SE supercomputing centre (Sweden). Potential ADME properties were investigated using QikProp from Maestro Suites 2015-2018 (Schrödinger Release 2018-4: QikProp, Schrödinger, LLC, New York, NY, 2018) as well as swissADME(15).

Quantitative realǦtime PCR - Total RNA was prepared using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA). All RNAs were reverseǦtranscribed with Maxima Reverse Transcriptase (Thermo Scientific, Waltham, MA, USA), according to

192 manufacturer protocol. qPCR was performed via a StepOnePlus™ RealǦTime PCR Systems from Applied Biosystems and the SYBR Green PCR Core reagents kit (TAKKARA). Analysis was carried out using QuantStudioTM Design and Analysis software version 1.3.1. Three technical repeats were performed per experiment. At least three biological repeats were performed per point per experiment. The primers used in the qPCR analyses were designed as follows: XBP1s forward primer (TGCTGAGTCCGCAGCAGGTG); XBP1s reverse primer (GCTGGCAGGCTCTGGGAAAG); Actin forward primer (CATGGGTGGAATCATAATGG); Actin reverse primer (AGCACTGTGTTGCGCTACAG).

Western blotting analyses - All IRE1 signalling analyses were carried out as described previously(16). Cells grown on 6 well plates were washed with PBS and lysed with RIPA lysis buffer at 4°C, over 25 minutes to extract protein. IRE1α and phosphorylated IRE1α were stained using antiǦIRE1α antibody (Anti-human; rabbit polyclonal; SantaCruz Biotechnologies, H-190) and pS724-IRE1 antibody (Anti- human; rabbit polyclonal; Abcam, ab48187). Cell extracts were resolved by SDS- PAGE and transferred to PVDF membranes. The resulting membranes were incubated with primary antibodies for 16 hrs at 4°C, washed with PBS, and incubated for 1 hour with goat antiǦrabbit secondary antibodies at room temperature (Invitrogen, Carlsbad, CA, USA) prior revelation using chemiluminescence.

193 Results Our previous studies indicated that large (18-50 amino acid long) peptides derived from the cytosolic domain of IRE1 could affect its oligomerisation and subsequent RNase activity(17). However, such peptides, even in their reduced 18 amino acid form, presented a plethora of issues such as bioavailability, proteasomal degradation, sheer size, crossing the blood brain barrier and stability when considering use in in vivo CNS settings. To address this, a library of overlapping tetra and penta-peptides derived from the IRE1 amino acid sequences previously identified to be biologically active (18) was generated in silico (Figure S1). This peptide library was then docked onto the ATP binding pocket of IRE1 using the Glide docking tool in Schrödinger(19) (Figure S1). Upon exploration of the binding pocket, peptides exhibiting superior docking scores were chosen for further modification and re-docked to obtain small peptides with maximal docking scores compared to ATP (PDB structure 3P23) or sulphonamide (PDB structure 4U6R) (Figure S1, S2, S6). Binding hypotheses based on pharmacophore modelling between the kinase domain and these strongly binding peptides were computationally generated (Figure S3). We then used these hypotheses to screen the 13000000 compound containing ZINC(20) database of clean lead-like compounds as well as a database of FDA approved molecules (ca. 1400 compounds) to yield structures with theoretically similar binding capabilities as the small peptides(20). The screen outcomes were then docked to the ATP binding pocket and the best scored candidates along with the best scoring peptides formed a final list of candidates to be further tested in vitro for their capacity to alter IRE1 RNase activity (Figure 1, Table1 and Figures S4, S5). The organic molecules thus identified include FDA approved compounds methotrexate, folinic acid, cefoperazone and fludarabine phosphate, and two molecules with no prior clinical designation, Z4 and Z6. All dockings were performed using standard precision (SP) and extra precision (XP)(21) and the OPLS-2005 force field(22). Methotrexate is used in chemotherapy against a variety of cancers and disease modifying treatment in a plethora of autoimmune disorders through two separate mechanisms. In cancer it serves as a folic acid antagonist(23) whilst in disorders such as rheumatoid arthritis it interferes with purine and pyrimidine precursors needed for cell proliferation(24). Folinic acid (or leucovorin) is co-prescribed with methotrexate to reduce methotrexate side effects(25). Cefoperazone is a third generation antibiotic from the cephalosporin family causing bacterial cell wall lysis

194 (26), whilst fludarabine phosphate is a chemotherapeutic used in haematological malignancies by disrupting DNA synthesis(27).

As exemplified by methotrexate in Figure 1B, the FDA approved compounds align well in the pocket with the nitrogen rich fused rings attaining the position of the adenine in ATP, and the oxygen rich ‘tails’ aligning along the phosphate binding region of the pocket. The similarity in interactions are also seen from the 2D interaction plots (Figures S9, S10). The Z4 and Z6 compounds are smaller in size to the FDA approved ones, both being nitrogen rich para substituted phenols. They interact with the ATP binding pocket through hydrogen bonding by the phenolic hydroxyl group to Cys645, an interaction also noted for nitrogens in the ‘adenine-like’ ring system in the FDA approved compounds, and hydrogen bonding between polar atoms at the opposite end of the molecules with Asp711 and Asp688 or Lys690 in the phosphate binding region of the pocket (Figure S8). Interestingly, of the different compounds obtained, it is only cefoperazone that was identified as a key binder from the ‘KIRA’-bound crystal structure 4U6R, which is slightly distorted relative to the ATP bound form; a distortion that has been claimed to be the key factor of the KIRA compounds ability to attenuate RNase activity. The remaining compounds were obtained as top hits upon binding to the ATP binding form 3P23, yet, as seen below, are nonetheless highly efficient in blocking RNase activity. This indicates that the mechanism of action of the KIRA compounds is slightly more complex than that the induced distortion affecting oligomerisation only.

The final list of candidates was tested in an in vitro cleavage assay as previously described (28) using MKC8866 as a positive control and DMSO as a negative control in the presence of recombinant IRE1, ATP, DTT, fluorescent Cy5-BHQ3 probe and test subject molecules. As with the in silico investigation, the peptides were tested first (Figure 2, Table 1 and Figure S11). Both F6P1 and F6P2 show some degree of IRE1 RNase inhibition (Figures 2A and S12). We then investigated the potential of these peptides to affect IRE1 function in a cell model. Increasing concentrations of each peptide were transfected in HEK293T cells over 4 hours and subsequently the cells were treated with 1μg/ml tunicamycin for a further 4 hours. XBP1 mRNA splicing was then tested as a marker of IRE1 RNase activity. As observed in vitro, both

195 peptides attenuated tunicamycin-induced IRE1-dependent XBP1 mRNA non- conventional splicing in HEK293T cells (Figure 2B).

We next tested the hits from the screened FDA approved drugs and small organic molecules in both primary (RADH87(12)) and common (U87) human GBM lines (Figure 3). Using MKC8866 as the control of RNase inhibition we monitored IRE1 activity in the presence of increasing concentrations of inhibitors. Interestingly in vitro all compounds showed a varying degree of inhibition (Figure 3A, S11, S13) confirming the hypothesis that the structure and binding based in silico screening yielded candidates with similar properties to the peptides. Since the majority of the compounds are therapeutics with known side effects, a toxicity screen was carried out in U87 and RADH87 GBM cells with increasing concentrations of each compound utilising a cell viability assay. IC50 was calculated for each compound and thus sub- toxic doses were determined for any and all subsequent assays for each cell line used (Figure 4A, 4B, S14). To verify the ability of the compounds to act as IRE1 inhibitors by impacting on IRE1 kinase activity, we treated U87 and RADH87 cells with sub-toxic doses of each compound for 24 hrs and measured the ratio of phospho-IRE1 (anti- P-S724) over total IRE1 compared to the untreated samples. The majority of our compounds reduced IRE1 auto-phosphorylation and hence potentially its activity at the protein level (Figure 3B, 3C, 3D, 3E). To test the ability of the identified compounds to inhibit IRE1 activity in human GBM cell models as well as determine the effect they may have upon induction of ER stress with different stressors, we treated U87 and RADH87 cells with sub-toxic doses of the FDA and non-FDA approved compounds in the presence of 1 μg/ml tunicamycin or 0.5 μg/ml SubAB toxin(29) for 24 hrs and 1.5 hrs respectively (Figure 3F,G,H,I). Tunicamycin is an antibiotic that causes ER stress via blocking N-linked protein glycosylation whilst SubAB acts as a serine protease cleaving BiP/GRP78(29), a chaperone essential for the recognition of unfolded protein load increase in the ER(30). As expected, the two ER stressors used herein showed different ER stress induction profiles, however, all compounds tested in the presence of tunicamycin or SubAB in U87 or RADH87 cells decreased XBP1 mRNA splicing over time. The induction profiles of both common (U87) and primary (RADH87(31)) GBM lines in the presence of tunicamycin were similar with the starkest differences seen at 16 and 24 hrs of treatment. ER stress induction with the SubAB toxin was much faster with maximal XBP1s expression

196 observed at 1.5 hrs after treatment and our test compounds took effect as early as 1 hr post administration. Such conserved effect of the molecules across different types of stress and cell lines solidifies the notion that compounds identified in our study are true inhibitors of IRE1 activity.

IRE1 has been shown to play a major role in GBM pathogenesis and pathophysiology both in animal models and human tumours(31). In mouse orthotopic brain models IRE1 deficient tumours displayed reduced angiogenesis and blood perfusion, growth and invasiveness(32), whilst double invalidation of both the RNase and kinase domains of IRE1 produced avascular tumours with blood vessel co-option and reduced macrophage infiltration capabilities(33). Moreover IRE1 mediates the control of GBM cell adhesion and migration through the cleavage of SPARC mRNA(34), and primary human GBM cells overexpressing an RNase deficient mutant form of IRE1 have shown severely impaired migration as well as monocyte infiltration of tumours(4). There have also been studies supporting the notion that induction of the UPR induces chemo-resistance in GBM(35) whilst high XBP1s activity has been suggested to be a negative prognostic marker of patient survival(4). It thus stands to reason that blocking IRE1 activity would be an attractive target in GBM treatment. However, when considering the clinical state of GBM or in fact any tumour in a systemic setting it is evident that sustained inhibition of IRE1 may have adverse rather than beneficial effects. After all IRE1 is part of a master homeostatic mechanism required by multiple systems for overcoming a diseased or “stressed” state. Consequently, a far more attractive hypothesis would be to sensitise tumours to the standard of care, chemotherapy and radiotherapy combination. In such a scenario tumour cells would be swiftly removed without the need for a prolonged exposure to IRE1 inhibition and potentially without prolonged exposure to highly toxic alkylating agents like TMZ.

To this end, we tested the effect of the IRE1 inhibiting compounds identified herein in our cellular GBM models when treated with TMZ. We first determined the toxicity of each molecule in U87 and RADH87 cells by treating cells with increasing concentrations of each molecule (0-1000 μM) over a period of 6 days upon completion of which cell survival was determined through WST1 mediated colorimetry and kill curve fit as described in the methods (Figure 4A, 4B). Guided by

197 these toxicity assays, cells were treated with sub-toxic doses of methotrexate, folinic acid, cefoperazone, fludarabine phosphate, Z4 and Z6 (Figure S14) and then subjected to increasing concentrations of TMZ (0-2500 μM) in the same temporal manner as with the toxicity investigation above. Kill curves were once again calculated (Figure 4C, D) and the IC50 of TMZ for U87 and RADH87 cells in the presence of each compound inferred (Figure 4E, 4F). All of the compounds tested significantly sensitised both commonly available and primary GBM cells to TMZ (Figure 4E, 4F).

Since the investigation yielded two novel, non-FDA approved, potential interventional candidates in the investigation of GBM adjuvant treatments, the projected ADME properties of these (Z4, Z6) were compared to those of an established IRE1 inhibitor (MKC8866) with particular focus on the blood brain barrier (BBB) to see whether they conferred an advantage over such inhibitor in CNS use. Using a dual in silico approach by utilising both QikProp from the Maestro Suite (Schrödinger Release 2018-4: Schrödinger, LLC, New York, NY, 2018) and the online tool swissADME(15), information was gathered concerning the compound lipophilicity (expressed as the logarithm of octanol/water partition coefficient), molecular weight, total polar surface, rotatable bonds and hydrogen bond donors, properties identified as important in the ability of a molecule to cross the BBB(36). These properties were compared to median values for each category observed in literature (Table 2). It has to be noted that these values cannot be taken into account individually as for example high lipophilicity may indeed increase the likelihood of a molecule to cross the BBB; however, it may also increase its capacity for non-specific binding and thus decrease its therapeutic potential(37).

198 Discussion Herein, we have thus uncovered a group of molecules that inhibit IRE1 activity and sensitise GBM cells (both established and primary lines) to TMZ chemotherapy when used at sub-toxic doses. Our investigation offers four major advantages to the pursuit of a solution to the temporally rapid progression and dismal end of GBM. Firstly, we here designed a drug discovery process by which already existing molecules that can inhibit IRE1 activity in silico, in vitro and in cell-based models can be identified. This, in an ever increasingly financially competitive environment of novel compound recognition, could prove an indispensable tool to bypass the need for new synthesis and thus fast-track drug repurposing and clinical benefit. Secondly, we identify novel modes of action for molecules already in abundant clinical use. Methotrexate for example is on the List of Essential Medicines(38) with dual mechanisms of action in various malignancies and auto-immune disorders. Its interference with IRE1 activity could very much impact its efficacy as a therapeutic as well as potentially open avenues of research into mechanisms of alleviating its side effects(39) as a UPR disruptor. Thirdly, we identify already clinically approved therapeutics that could potentially have a beneficial effect on GBM; a disease in dire need for improved clinical options. Such discoveries are of great benefit to clinical trial design as they can be put straight into the clinic with much reduced temporal and financial burden onto the design of the trial itself. This means that patients could benefit from such insights in an expedient manner. Lastly, we define an approach to test molecules that have no clinical designation yet. Compounds Z4 and Z6 are both excellent candidates for use in GBM. Z6 greatly sensitises GBM lines to TMZ whilst Z4 although as efficient as Z6 at synergising with TMZ, has a much more favourable toxicity profile with doses of up to 1000 μM being non-toxic across both common and primary GBM lines. This offers the advantage of reduced potential for off target toxicity. Moreover, in the context of drug delivery through the BBB, Z4 has more favourable ADME properties when compared to MKC8866 or Z6 as it more closely conforms to the consensus values for CNS drug design (Table 2). This of course does not stop Z6 being a valid candidate as alternative methods for delivery into the brain to treat CNS malignancies have been described(40) and sensitisation of tumours to chemotherapy may well benefit

199 from intra-operative administration without requiring repeat intravenous or oral treatment.

In conclusion through the structural interrogation of the IRE1 kinase catalytic domain and subsequent in silico screen we have identified four clinically approved molecules and two novel, hitherto not clinically approved molecules, that inhibit IRE1 RNase activity in vitro and in human GBM cell models, and sensitise them to the established GBM chemotherapy treatment, temozolomide.

200 Acknowledgements We thank Chetan Chintha and Afshin Samali (National University of Ireland Galway, Ireland) for constructive discussion. This work was supported by grants from Institut National du Cancer (INCa; PLBIO), Agence Nationale de la Recherche (ANR; ERAAT) and Fondation pour la Recherche Médicale (FRM; Equipe labellisée 2018) to EC. DD, CC, AS, LAE and EC are funded by EU H2020 MSCA ITN-675448 (TRAINERS) and MSCA RISE-734749 (INSPIRED) grants. The Swedish Research Council (VR) and the Swedish National Infrastructure for Computing (SNIC) are gratefully acknowledged for funding and allocations of computing time, respectively (LAE).

201 References 1. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJB, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005;352:987–96. 2. Doultsinos D, Avril T, Lhomond S, Dejeans N, Guédat P, Chevet E. Control of the Unfolded Protein Response in Health and Disease. SLAS Discov Adv Life Sci R&D. SAGE Publications Inc STM; 2017;2472555217701685. 3. Obacz J, Avril T, Le Reste P-J, Urra H, Quillien V, Hetz C, et al. Endoplasmic reticulum proteostasis in glioblastoma—From molecular mechanisms to therapeutic perspectives. Sci Signal. 2017;10. 4. Lhomond S, Avril T, Dejeans N, Voutetakis K, Doultsinos D, McMahon M, et al. Dual IRE1 RNase functions dictate glioblastoma development. EMBO Mol Med. 2018;e7929. 5. Logue SE, McGrath EP, Cleary P, Greene S, Mnich K, Almanza A, et al. Inhibition of IRE1 RNase activity modulates the tumor cell secretome and enhances response to chemotherapy. Nat Commun. 2018;9:3267. 6. Zhao N, Cao J, Xu L, Tang Q, Dobrolecki LE, Lv X, et al. Pharmacological targeting of MYC-regulated IRE1/XBP1 pathway suppresses MYC-driven breast cancer. J Clin Invest. The American Society for Clinical Investigation; 2018;128:1283–99. 7. Hetz C, Chevet E, Harding HP. Targeting the unfolded protein response in disease. Nat Rev Drug Discov. 2013/08/31. 2013;12:703–19. 8. Wang L, Perera BGK, Hari SB, Bhhatarai B, Backes BJ, Seeliger MA, et al. Divergent allosteric control of the IRE1α endoribonuclease using kinase inhibitors. Nat Chem Biol. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.; 2012;8:982–9. 9. Feldman HC, Tong M, Wang L, Meza-Acevedo R, Gobillot TA, Lebedev I, et al. Structural and Functional Analysis of the Allosteric Inhibition of IRE1α with ATP-Competitive Ligands. ACS Chem Biol. 2016;11:2195–205. 10. Bouchecareilh M, Higa A, Fribourg S, Moenner M, Chevet E. Peptides derived from the bifunctional kinase/RNase enzyme IRE1α modulate IRE1α activity and protect cells from endoplasmic reticulum stress. FASEB J. 2011;25:3115– 29. 11. Bouchecareilh M, Caruso M-E, Roby P, Parent S, Rouleau N, Taouji S, et al.

202 AlphaScreen-based characterization of the bifunctional kinase/RNase IRE1alpha: a novel and atypical drug target. J Biomol Screen. 2010;15:406–17. 12. Avril T, Etcheverry A, Pineau R, Obacz J, Jegou G, Jouan F, et al. CD90 Expression Controls Migration and Predicts Dasatinib Response in Glioblastoma. Clin Cancer Res. 2017;23:7360 LP-7374. 13. Goede A, Michalsky E, Schmidt U, Preissner R. SuperMimic--fitting peptide mimetics into protein structures. BMC Bioinformatics. BioMed Central,; 2006;7:11. 14. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, et al. UCSF Chimera - A visualization system for exploratory research and analysis. J Comput Chem. 2004;25:1605–12. 15. Daina A, Michielin O, Zoete V. SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci Rep. Nature Publishing Group; 2017;7:42717. 16. Lhomond S, Pallares N, Barroso K, Schmit K, Dejeans N, Fazli H, et al. Adaptation of the secretory pathway in cancer through IRE1 signaling. Methods Mol Biol. 2015/03/26. 2015;1292:177–94. 17. Bouchecareilh M, Higa A, Fribourg S, Moenner M, Chevet E. Peptides derived from the bifunctional kinase/RNase enzyme IRE1 modulate IRE1 activity and protect cells from endoplasmic reticulum stress. FASEB J. 2011;25:3115–29. 18. Bouchecareilh M, Higa A, Fribourg S, Moenner M, Chevet E. Peptides derived from the bifunctional kinase/RNase enzyme IRE1{alpha} modulate IRE1{alpha} activity and protect cells from endoplasmic reticulum stress. FASEB J. 2011;25:3115–29. 19. Halgren TA, Murphy RB, Friesner RA, Beard HS, Frye LL, Pollard WT, et al. Glide: A New Approach for Rapid, Accurate Docking and Scoring. 2. Enrichment Factors in Database Screening. J Med Chem. American Chemical Society; 2004;47:1750–9. 20. Sterling T, Irwin JJ. ZINC 15--Ligand Discovery for Everyone. J Chem Inf Model. 2015/10/19. American Chemical Society; 2015;55:2324–37. 21. Friesner RA, Murphy RB, Repasky MP, Frye LL, Greenwood JR, Halgren TA, et al. Extra Precision Glide: Docking and Scoring Incorporating a Model of Hydrophobic Enclosure for Protein−Ligand Complexes. J Med Chem. American Chemical Society; 2006;49:6177–96.

203 22. Banks JL, Beard HS, Cao Y, Cho AE, Damm W, Farid R, et al. Integrated Modeling Program, Applied Chemical Theory (IMPACT). J Comput Chem. 2005;26:1752–80. 23. Farber S, Diamond LK, Mercer RD, Sylvester RF, Wolff JA. Temporary Remissions in Acute Leukemia in Children Produced by Folic Acid Antagonist, 4-Aminopteroyl-Glutamic Acid (Aminopterin). N Engl J Med. Massachusetts Medical Society; 1948;238:787–93. 24. Cutolo M, Sulli A, Pizzorni C, Seriolo B, Straub RH. Anti-inflammatory mechanisms of methotrexate in rheumatoid arthritis. Ann Rheum Dis. 2001;60:729 LP-735. 25. Shiroky JB, Neville C, Esdaile JM, Choquette D, Zummer M, Hazeltine M, et al. Low-dose methotrexate with leucovorin (folinic acid) in the management of rheumatoid arthritis. results of a multicenter randomized, double-blind, placebo- controlled trial. Arthritis Rheum. John Wiley & Sons, Ltd; 2018;36:795–803. 26. Jones RN, Barry AL. Cefoperazone: A Review of Its Antimicrobial Spectrum, β- Lactamase Stability, Enzyme Inhibition, and Other in Vitro Characteristics. Rev Infect Dis. Oxford University Press; 1983;5:S108–26. 27. Keating MJ, Kantarjian H, Talpaz M, Redman J, Koller C, Barlogie B, et al. Fludarabine: a new agent with major activity against chronic lymphocytic leukemia. Blood. 1989;74:19 LP-25. 28. Volkmann K, Lucas JL, Vuga D, Wang X, Brumm D, Stiles C, et al. Potent and selective inhibitors of the inositol-requiring enzyme 1 endoribonuclease. J Biol Chem. 2011/02/10. 2011;286:12743–55. 29. Wolfson JJ, May KL, Thorpe CM, Jandhyala DM, Paton JC, Paton AW. Subtilase cytotoxin activates PERK, IRE1 and ATF6 endoplasmic reticulum stress-signalling pathways. Cell Microbiol. 2008/04/21. 2008;10:1775–86. 30. Bertolotti A, Zhang Y, Hendershot LM, Harding HP, Ron D. Dynamic interaction of BiP and ER stress transducers in the unfolded-protein response. Nat Cell Biol. 2000/06/15. 2000;2:326–32. 31. Lhomond S, Avril T, Dejeans N, Voutetakis K, Doultsinos D, McMahon M, et al. Dual IRE1 RNase functions dictate glioblastoma development. EMBO Mol Med. 2018;10:139–308. 32. Auf G, Jabouille A, Guerit S, Pineau R, Delugin M, Bouchecareilh M, et al. Inositol-requiring enzyme 1alpha is a key regulator of angiogenesis and

204 invasion in malignant glioma. Proc Natl Acad Sci U S A. 2010/08/13. 2010;107:15553–8. 33. Jabouille A, Delugin M, Pineau R, Dubrac A, Soulet F, Lhomond S, et al. Glioblastoma invasion and cooption depend on IRE1α endoribonuclease activity. Oncotarget. 2015;6:24922–34. 34. Dejeans N, Pluquet O, Lhomond S, Grise F, Bouchecareilh M, Juin A, et al. Autocrine control of glioma cells adhesion and migration through IRE1alpha- mediated cleavage of SPARC mRNA. J Cell Sci. 2012/06/22. 2012;125:4278– 87. 35. Epple LM, Dodd RD, Merz AL, Dechkovskaia AM, Herring M, Winston BA, et al. Induction of the unfolded protein response drives enhanced metabolism and chemoresistance in glioma cells. PLoS One. 2013/09/17. 2013;8:e73267. 36. Pajouhesh H, Lenz GR. Medicinal chemical properties of successful central nervous system drugs. NeuroRx. The American Society for Experimental neuroTherapeutics, Inc.; 2005;2:541–53. 37. Rankovic Z. CNS Drug Design: Balancing Physicochemical Properties for Optimal Brain Exposure. J Med Chem. American Chemical Society; 2015;58:2584–608. 38. WHO | WHO Model Lists of Essential Medicines [Internet]. WHO. World Health Organization; 2018 [cited 2018 Nov 9]. Available from: http://www.who.int/medicines/publications/essentialmedicines/en/ 39. Gibson EM, Nagaraja S, Ocampo A, Tam LT, Wood LS, Pallegar PN, et al. Methotrexate Chemotherapy Induces Persistent Tri-glial Dysregulation that Underlies Chemotherapy-Related Cognitive Impairment. Cell. Elsevier; 2018; 40. Carpentier A, Canney M, Vignot A, Reina V, Beccaria K, Horodyckid C, et al. Clinical trial of blood-brain barrier disruption by pulsed ultrasound. Sci Transl Med. 2016;8. 41. Ali MM, Bagratuni T, Davenport EL, Nowak PR, Silva-Santisteban MC, Hardcastle A, et al. Structure of the Ire1 autophosphorylation complex and implications for the unfolded protein response. EMBO J. 2011/02/15. 2011;30:894–905. 42. Harrington PE, Biswas K, Malwitz D, Tasker AS, Mohr C, Andrews KL, et al. Unfolded Protein Response in Cancer: IRE1α Inhibition by Selective Kinase Ligands Does Not Impair Tumor Cell Viability. ACS Med Chem Lett. American

205 Chemical Society; 2014;6:68–72. 43. Wager TT, Chandrasekaran RY, Hou X, Troutman MD, Verhoest PR, Villalobos A, et al. Defining desirable central nervous system drug space through the alignment of molecular properties, in vitro ADME, and safety attributes. ACS Chem Neurosci. American Chemical Society; 2010;1:420–34. 44. Wager TT, Hou X, Verhoest PR, Villalobos A. Moving beyond rules: the development of a central nervous system multiparameter optimization (CNS MPO) approach to enable alignment of druglike properties. ACS Chem Neurosci. American Chemical Society; 2010;1:435–49.

206 Table 1. A detailed representation of in silico and in vitro data for each compound yielded by the screen as described in figure 1. Data are presented for compounds used as controls for the in silico docking such as ATP for IRE1 crystal PDB structures 3P23(41) and sulphonamide for PDB structure 4U6R(42). All compounds obtained from the in silico small peptide design or ZINC database screen were docked onto both receptors and the scores are presented here. The calculated in vitro IC50s are extrapolated from percentage inhibition of IRE1 activity curves as shown in figures 2A and 3A. The final four columns are dedicated to toxicity IC50s for U87 and RADH87 GBM cell lines as shown in figure 3C & D, and U87 and RADH87 sensitisation to TMZ in the presence of non-toxic doses of each compound (Figure S14).

207 Kinase IRE1 Docking In vitro Toxicity (μM) TMZ IC50 (μM) Compound Compound type Clinical Use Receptor Target score IC50 (μM) (PDB) U87 RADH87 U87 RADH87 ATP 3P23 -8.044 N/A N/A N/A N/A N/A N/A Sulphonamide 4U6R -7.815 Control MKC8866 N/A N/A N/A 0.431 N/A N/A N/A N/A

Temozolomide Chemotherapy N/A N/A N/A N/A N/A >2500 >2500

3P23 -9.442 F6P1 Peptide N/A Kinase 2.8031 N/A N/A N/A N/A 4U6R -7.783 3P23 -9.260 F6P2 Peptide N/A Kinase 16.322 N/A N/A N/A N/A 4U6R -9.125

3P23 -5.211

208 Cefoperazone Cephalosporin Bacterial Infection Kinase 0.164 >1000 996.0 1786 1225 4U6R -8.090

3P23 -9.478 Methotrexate Anti-folate Chemotherapy Kinase 11.173 >1000 132.4 1957 2309 4U6R -4.323

3P23 -8.838 Folinic Acid Folic acid Derivative Chemotherapy Kinase 0.2298 >1000 >1000 2124 2824 4U6R -5.546

Fludarabine 3P23 -8.188 Purine Analogue Chemotherapy Kinase 0.114 27.03 66.59 2434 1850 Phosphate 4U6R -6.865

3P23 -8.68 Z4 Unspecified N/A Kinase 21.736 >1000 >1000 2401 1950 4U6R -5.476

3P23 -8.423 Z6 Unspecified N/A Kinase 19.753 63.10 150.6 658.6 839.6 4U6R -3.547 Table 2. A representation of projected ADME property values for non FDA approved compounds Z4, Z6 and IRE1 ribonuclease inhibitor MKC8866. The values included in the table were chosen as the most important indicators of blood brain barrier crossing capabilities according to literature(37).

Median values for Z4 Z6 MKC8866 ADME marketed CNS Property drugs(36,43,44) QikProp swissADME QikProp swissADME QikProp swissADME

Log Po/w ≤3 1.767 2.57 0.403 0.52 1.185 1.97

Hydrogen <3.5 5 4 3 301 Bond Donors

Topological Polar Surface 40

Rotatable <8 7 8 4 398 bonds

Molecular ≤360 341.40 275.31 336.34 Weight (Da)

209 Figure Legends

Figure 1. A) Pipeline of candidate library discovery from in silico peptide test subject design to the kinase ATP binding pocket exploration, ZINC database screen, and final candidate library. B) Examples of docking of candidates in the ATP binding pocket of IRE1 (3P23 PDB structure) showing ATP as the control, F6P1 as a peptide example and methotrexate as an FDA approved therapeutic. C) F6P1 (tetra-peptide) and F6P2 (penta-peptide) derived from the kinase domain sequence of IRE1. D), E) FDA approved and non FDA approved molecules, respectively, with docking scores for the kinase active site similar to or better than ATP.

Figure 2. A) Activity of IRE1 in the presence of F6P1 and F6P2 compared to MKC8866 stemming from FRET signal given out upon fluorescent probe cleavage over 25 minute incubation. B) HEK293T cells were transfected with increasing concentrations of F6P1 or F6P2 and treated with tunicamycin. XBP1s mRNA levels were measured after a 4 hr treatment as a measure of IRE1 RNase activity

Figure 3. Nomenclature and symbols for compounds are introduced at the top left hand corner of the figure. A) Activity of IRE1 in the presence of methotrexate, folinic acid, cefoperazone, fludarabine phosphate, Z4 and Z6 compared to MKC8866 stemming from FRET signal given out upon fluorescent probe cleavage over 25 minute incubation. B, C) Representative blot of U87 and RADH87 cell treatment blotted for IRE1 and phospo-IRE1. D, E) Protein levels of IRE1 and phospho-IRE1 in U87 (D) and RADH87 (E) cells treated with methotrexate, folinic acid, cefoperazone, Z4 and Z6 over 24 hours. Fold change of protein expression between IRE1 and phospho-IRE1 is represented in bar chart form normalised to untreated U87 (D) and untreated RADH87 (E) cells. F, G) XBP1s mRNA levels upon treatment with 1μg/ml Tunicamycin over 24hrs normalised to untreated U87 (F) and RADH87 (G) cells. H, I) XBP1s mRNA levels upon treatment with 0.5 μg/ml SubAB toxin over 1.5hrs normalised to untreated U87 (H) and RADH87 (I) cells.

Figure 4. Nomenclature and symbols for compounds are introduced at the bottom right hand corner of the figure. A, B) Percent cell viability measured via WST1 mediated colourimetry in the presence of increasing concentrations of IRE1 inhibitors in U87 (A) and RADH87 (B) cells. C, D) Percent cell viability tested by colorimetric

210 WST1 assay upon treatment with increasing concentrations of TMZ and non-toxic concentrations of methotrexate, folinic acid, cefoperazone, fludarabine phosphate, Z4 and Z6 in U87 (C) and RADH87 (D) cells. E, F) TMZ IC50s in the presence of non- toxic doses of IRE1 inhibitors in U87 (E) and RADH87 (F) cells

211 Figure 1

212 Figure 2

213 Figure 3

214 Figure 4

215 SUPPLEMENTAL INFORMATION Supplemental Figures

Figure S1: Identification of IRE1 activity modulator candidate library. A) Cytosolic IRE1 structure. As previously described[1], fragments of cytosolic IRE1 were shown to impact IRE1 activity. The amino acid sequences of these fragments (F6) were dissected in sequential tetra and pentapeptide fragments to B) produce a library of small peptides. C) These small peptides were computationally docked onto the corresponding proposed IRE1 sites of action, determined by the sequence of the fragments themselves (eg F6 derived fragments would dock onto the kinase domain). D) The best candidates (determined by docking score) were modified using structural computational approaches to produce a series of peptidomimetics that mimic the small peptide structure. Moreover the docking hypotheses of these molecules were screened against the ZINC/FDA approved molecule databases to produce a series of small molecules that could mimic the small peptide binding on IRE1. E) These were docked on IRE1 as in C). This approach yielded a final library of small peptides, peptidomimetics and organic molecules to be tested in vitro and in vivo.

216

Figure S2: Structures utilised in docking. In order to produce as accurate representations of potential IRE1 kinase binders, two crystal IRE1 structures were used available in the protein data bank (PDB). A) The 3P23 structure was used as a template for molecules that would bind IRE1 in a conformation similar to this of ADP[2]. B) The 4U6R structure was used as a template for molecules that would bind IRE1 in a conformation similar to this of a sulphonamide[3]. C) The two structures superimposed to illustrate the differences in the shape of the kinase pocket when ADP (smaller molecule) or sulphonamide (larger molecule) is bound.

217

Figure S3: Peptide pharmacophore hypotheses. Pharmacophore hypotheses were generated for both F6P1 and F6P2 on both 3P23 and 4U6R using receptor-ligand complex modelling. A) F6P1 pharmacophore model based on interaction with 3P23 kinase pocket. B) F6P1 pharmacophore model based on interaction with 4U6R kinase pocket. C) F6P2 pharmacophore model based on interaction with 3P23 kinase pocket. D) F6P2 pharmacophore model based on interaction with 4U6R kinase pocket.

218 Figure S4: F6P1 binding hypothesis screen docked on 3P23. A screen of ZINC and FDA approved databases with the F6P1-3P23 ligand-receptor hypothesis yielded a combined 3687 hits which were docked onto the 3P23 structure. Docking scores varied from -9.478 to -0.34 and representative scores were chosen to illustrate this range. The threshold value for hit selection was set by the docking score of ATP. In light grey are compounds that were above the threshold but were either unavailable for purchase or were simply ATP analogues. In black are compounds that were below the threshold. In light pink and light green are non FDA approved compounds and FDA approved compounds respectively that were not selected due to the closeness to the threshold. In dark reds and dark greens (colour codes used in the main manuscript) are the FDA and non FDA approved compounds selected for further experimentation.

219 Figure S5: F6P2 binding hypothesis screen docked on 4U6R. A screen of ZINC and FDA approved databases with the F6P1-3P23 ligand-receptor hypothesis yielded a combined 2706 hits which were docked onto the 4U6R structure. Docking scores varied from -8.188 to -0.162 and representative scores were chosen to illustrate this range. The threshold value for hit selection was set by the docking score of Sulphonamide. In light grey are compounds that were above the threshold but were either unavailable for purchase or were simply ATP analogues. In black are compounds that were below the threshold. In dark greens are the FDA approved compounds chosen for further experimentation.

220 Figure S6: F6P1 Ligand interaction map. A) Ligand interaction map of F6P1 and the 3P23 ATP kinase binding pocket. B) Ligand interaction map of F6P1 and the 4U6R ATP kinase binding pocket.

221 Figure S7: F6P2 Ligand interaction map. A) Ligand interaction map of F6P2 and the 3P23 ATP kinase binding pocket. B) Ligand interaction map of F6P2 and the 4U6R ATP kinase binding pocket.

222

Figure S8: Z4 and Z6 ligand interaction maps. A) Ligand interaction map of Z4 and the 3P23 ATP kinase binding pocket. B) Ligand interaction map of Z6 and the 3P23 ATP kinase binding pocket.

223 Figure S9: Methotrexate and Folinic Acid ligand interaction maps. A) Ligand interaction map of Methotrexate and the 3P23 ATP kinase binding pocket. B) Ligand interaction map of Folinic Acid and the 3P23 ATP kinase binding pocket.

224 Figure S10: Cefoperazone and Fludarabine phosphate ligand interaction maps. A) Ligand interaction map of Cefoperazone and the 4U6R ATP kinase binding pocket. B) Ligand interaction map of Fludarabine and the 3P23 ATP kinase binding pocket.

225

Figure S11: Fluorescence cleavage assay design and validation. A) An RNA probe was designed to include a 3’ Cy5 fluorescent dye and a 5’ BHQ3 quencher. B) Upon introducing recombinant IRE1, the probe is cleaved leading to an increase in fluorescence. C) Assay validation in the presence or absence of recombinant IRE1.

226 Figure S12: In vitro cleavage assay fluorescence curves. A-C) Representative curves of varying concentrations of molecules and control (DMSO for FDA and non FDA approved molecules and reaction buffer for peptides). The key to the figure can be seen at the bottom right.

227 Figure S13: In vitro cleavage assay fluorescence curves. A-G) Representative curves of varying concentrations of molecules and control (DMSO for FDA and non FDA approved molecules). MKC8866 was used as a comparator for efficacy of RNAse activity blocking. The key to the figure can be seen at the bottom right.

228

Figure S14: Molecule cell toxicity. A-B) Toxicity IC50s for each inhibitor in U87 and RADH87 cells. C) Concentrations of each inhibitor used in each cell line for each separate investigation.

229 References [1] M. Bouchecareilh, A. Higa, S. Fribourg, M. Moenner, E. Chevet, FASEB J 2011, 25, 3115– 3129. [2] M. M. Ali, T. Bagratuni, E. L. Davenport, P. R. Nowak, M. C. Silva-Santisteban, A. Hardcastle, C. McAndrews, M. G. Rowlands, G. J. Morgan, W. Aherne, et al., EMBO J 2011, 30, 894–905. [3] P. E. Harrington, K. Biswas, D. Malwitz, A. S. Tasker, C. Mohr, K. L. Andrews, K. Dellamaggiore, R. Kendall, H. Beckmann, P. Jaeckel, et al., ACS Med. Chem. Lett. 2015, 6, 68–72. [4] T. Avril, A. Etcheverry, R. Pineau, J. Obacz, G. Jegou, F. Jouan, P.-J. Le Reste, M. Hatami, R. R. Colen, B. L. Carlson, et al., Clin. Cancer Res. 2017, 23, 7360 LP-7374.

230 Chapter 6: Conclusions and Contributions

In this sixth chapter we have structurally interrogated the IRE1 kinase catalytic domain and subsequently carried out an in silico screen that identified four clinically approved and two novel, not clinically approved molecules that inhibit IRE1 RNase activity in vitro and in common as well as primary GBM cells. Following this the ability of GBM cells to cope with TMZ in the presence of these molecules was tested, providing evidence that IRE1 kinase inhibition sensitises human GBM cell lines to TMZ. We have thus overseen the completion of the third primary objective by producing novel molecules that sensitise GBM cells to current therapies.

The author of this thesis exclusively contributed to the composition of “Novel IRE1 inhibitors chemo-sensitise Glioblastoma cells to Temozolomide” both in writing of the manuscript and carrying out both the computational13 and biological experimental work14.

Particular thanks to Prof. L. A. Eriksson for making the commencement and realisation of a project outside this author’s expertise both feasible and familiar but also enjoyable and productive.

13 Special thanks to A. Carlesso for training in using specific computational chemistry software and to Dr E. Chevet and Professor L. A. Eriksson for continuing advice and input in the design and implementation of the study. 14 Special thanks to Professors A. and J. Paton who supplied the SubAb toxin for the completion of this study.

231 Discussion

We set out at the beginning of this thesis as naïve observers in the field of ER protein homeostasis (a.k.a. proteostasis) and IRE1 signalling with a purpose to find justification in dedicating valuable resources to targeting IRE1 in the field of neuro-oncology. We have since explored the ER structure and familiarised with its functions and extrinsic and intrinsic perturbations, detailing the effects and consequences of ER stress. The UPR was introduced as a major mechanism through which this stress is dealt with and IRE1 was first introduced as an important transducer of the UPR. The UPR’s importance in inter-organelle and inter-organ communication was defined as well as its role in cellular life and death decisions in normal physiology. ER stress is central to life itself as without it embryology and development, immunity and growth would be non-existent. It came as no surprise then that if coping mechanisms of these signalling pathways were to be overwhelmed a multitude of issues and diseases would occur. Naturally such integral cogs in the machine of life have been touted as pharmacological targets with efforts to bring the UPR in the clinic continuous and numerous but without immense success. One has to wonder at why and of course specificity may well be at the centre of such a meagre return of targeting success. UPR signalling is a molecular cluster of interconnected players constantly oscillating around a norm; an impeccable example of homeostasis. If indeed one attempts to stifle its activity or indeed boost it in diseased tissue then at the same time they would be introducing stress inducing oscillations elsewhere in the intricately interconnected universe that is the human body. Such a consideration is important in the field of translational research as its target is not only to produce high quality research but also to produce clinically relevant information that will do no harm to the patients that may benefit from it.

We then explored the complexities of the discovery processes of molecules that can impact the UPR and appreciated the multidisciplinary nature of the work that needs to be undertaken for a molecule to be picked from a myriad of others to be used in research let alone the clinic. Having familiarised with the general concept of the UPR we started to focus on a particular player in the ER stress response cascade; IRE1, the most evolutionarily conserved UPR transducer. Once again it became quite clear that IRE1 was of utmost importance in multiple processes playing contrasting roles in neurodegeneration where it promoted neuronal death whilst it aided glioblastoma cells to adapt to an increasingly hostile microenvironment saturated

232 with hypoxic and acidic factors before even considering the toxicity that chemotherapeutics could convey.

Reflecting upon the wide reaching consequences of IRE1 biology we were confronted with a tremendous clinical problem. A cancer quite simply without match, which brings debilitating symptoms to patients that can only hope for a few months of life after their diagnosis is known. For almost 200 years GBM has been characterised, researched and combated with a multitude of tools and techniques with a clear answer yet to present itself.

What are we then left with? A molecular pathway integral to the life of a cell and patients that are in dire need for a new treatment that could if not cure, at least give them a little bit more time or a better quality of life during their remaining days. There is thus, this author feels, a strong rationale in targeting IRE1 in GBM, not blindly using this molecular network as a therapeutic target but responsibly and methodically gathering information as to its exact involvement in how the disease behaves and then attempt to better the status quo by synergising with current therapeutic options to provide realistic clinical outcomes through a preclinical investigation.

So how to carry out such an investigation? In science or medicine, on a lab bench or an operating theatre the physician or scientist always first takes stock of what is available to them be that knowledge or materials and deduces what is needed based on the questions that need answering. This is the very approach that has been taken during this investigation.

“IRE1 is a major pathophysiological mediator and valid pharmacological target in GBM and its modulation may provide novel therapeutic options as an adjuvant disease modifying treatment.”

A translational model of altered IRE1 activity in GBM

This was the hypothesis underlining this thesis. What we had was clues as to the importance of IRE1 in GBM oncogenesis. What we wanted to deduce was the extent to which this was true and whether it was worth and safe to pharmacologically target IRE1 in GBM. Studies prior to the ones presented in here had demonstrated the importance of IRE1 in tumour aggressiveness but they did not provide the underlying molecular mechanisms of this effect. Moreover previous studies did not recapitulate the brain tumour complexity when considering genetic signatures of IRE1 GBM involvement. What we thus needed was a working experimental model to

233 accompany the hypothesis that recapitulated a patient population. As such this model had to be human. The use of animals in medical research has been absolutely indispensable however, one cannot circumvent their biggest flaw: quite simply they are not human. Moreover the model had to recapitulate a trait found everywhere in nature: inter-individual variability and heterogeneity. Finally, it had to mimic characteristics presented by human GBM tumours, relevant to IRE1 functionality. Where better to obtain such a combination of traits that the source itself. Short of growing tumours in vitro, deriving cell lines from a large range of resected patient tumours is the closest one could get to a humanised experimental model that can be readily used in a laboratory setting. There are of course limitations since physiological and anatomical features cannot be recapitulated especially when considering a surgical model. However, primary cells lines hold distinct advantages over commonly available lines in that they retain many of the genotypic and phenotypic characteristics of the original tissue whilst common cell lines may vary greatly.

The transcriptomic characterisation of these primary lines formed the backbone of their validation as an experimental model. At the start of chapter 4 we overlooked the analysis of a large dataset (TCGA) when confronted with an IRE1 –dependent gene expression signature to determine the effect of IRE1 signalling from a transcriptomic basis on patient survival and tumour properties. This was done using a bioinformatics pipeline termed BioInfominer15 that used a series of algorithms to combine signalling pathway prioritisation based on gene ontology and graph theory analysis to come up with a list of genes prioritised for their relevance to IRE1 function. This was then confronted to an IRE1 signalling signature described in a previous publication to come up with a final list of 38 genes that could serve as a beacon of IRE1 activity in transcriptomic datasets. This interaction highlights the importance of big data mining in medical research. There is little point in accumulating riches of information without having the power to analyse it to extract meaningful deductions; as such we verify that confronting a disease requires much more than clinicians or biological scientist but rather a harmonious collaboration of a multitude of clusters of individuals; a concept not easy to achieve as it may be hampered by sheer distances or lack of understanding of requirements and timescales between different fields of study. Such occupational hazards are often overlooked when considering the attrition rates of novel treatment development but they are critical in the successful completion of an investigation. As such it is not just scientific or clinical data that are important deliverables of an investigation, rather inter- and intra-group communication and effective collaboration is an acquired skill that is indispensable in disease prevention.

15 This work was carried out in collaboration with Prof A. Chatziioannou (NHRF, Greece) and the pipeline is characterised in: Lhomond et al; EMBO Mol. Med. 2018 and

234 The next step was to deduce any “naturally occurring” impact that IRE1 may have on GBM and to do this we searched for mutations in IRE1 in such tumours. By uncovering four distinct variants we were able to speculate on the potential outcomes of such mutations in the structural integrity of IRE1 and subsequently on the outcomes of its enzymatic activity. After all, the aim of this investigation was to uncover the underlying mechanisms of IRE1 oncogenicity In GBM. Indeed each variant conferred characteristics that enabled the characterisation of distinct phenotypes allowing for the notion of a dual IRE1 role in oncogenesis to emerge. The XBP1s signalling axis promoted immune cell tumour infiltration, increased angiogenesis and the expression of migration and invasion markers whilst RIDD activity against mRNA or miRNA reduced the effect of angiogenesis as well as migration and invasion. What this showed is that IRE1 may well be a rheostat in the true sense of the word, a molecule that can work in both directions conferring both beneficial and detrimental characteristics to a tumour according to its needs. Of course such signalling ambivalence does not only provide tumours with a varied repertoire to survive and grow but also provides clues as to the targeting of such pathways to exploit this balance for therapeutic purposes.

Therefore to test such a potential primary cell lines were extracted from resected tumours and not only provided cells that conferred the advantages of tissue specificity as described earlier but also due to their selection using different growth media, provided two distinct phenotypes derived from each tumour. As such the resulting lines would represent both a cancer stem cell-like neurosphere phenotype and a cancer differentiated cell-like adherent phenotype; characteristics that proved of great significance in chapter 5. These lines were scrutinised for their transcriptomic compatibility with the patient data available and indeed they clustered in groups of IRE1 activity depending on their XBP1s or RIDD status mimicking the clusters seen in patient groups and they showed phenotypic variability that recapitulated tumour heterogeneity. To produce an experimental model of altered IRE1 activity in such lines we needed to consider the possible desirable outcomes of any such manipulation as we had shown that not only did these lines cluster in XBP1/RIDD groups of varying levels but also that each of these pathways may have differential outcomes on GBM development. It was thus prudent to not just consider ablating IRE1 altogether as such a model would lose its underlying fine-tuning details of variant IRE1 activity. In order to cover a complete range of cellular tools lines were created overexpressing WT or mutant forms of IRE1 found in GBM tumours in four primary lines that segregated to different XBP1/RIDD groups as dictated by patient data. The mutants comprised Q780* and S769F, a truncation mutant that lacked a ribonucleases domain and a mutant with a substitution mutation likely leading to a conformational change in the quaternary structure of IRE1 affecting its function, respectively. By producing these lines and

235 subsequently showing that they had distinct biological outcomes concerning XBP1 splicing, UPR marker expression, EMT marker expression, chemokine expression, tumour cell migration and monocyte migration to the tumour it is evident that we have a validated translational cellular model of altered IRE1 activity in GBM to further test the validity of IRE1 as a target in GBM. To enhance this model more primary lines expressing the WT and mutant forms of IRE1 could be generated and fully characterised. This way a predictor of various treatment options would be generated in lines recapitulating the variable XBP1/RIDD output observed in patients. Moreover characterising the S796F mutant fully using in silico computational methods would be of great interest to see what effect this mutation could have on the final quaternary structure of IRE1 and hence what role it may play in its output when expressed in lines predisposed to RIDD or XBP1s signalling. Such manipulation would provide more information on the effect of targeting IRE1 without completely ablating its RNase domain would have on GBM cells and would thus complement and aid the investigations in novel therapeutics and drug resistance discussed below.

From an IRE1 centric investigation in patient, human cell and animal model we documented key regulatory mechanisms of GBM development and produced the tools to test them further. However, one other major theme that arose from this investigation was the breadth of heterogeneity of IRE1 expression and molecular output in GBM tumours. When considering it as a therapeutic target, it would surely be irresponsible not to take into account the possibilities of potential target efficacy as well as off target toxicity whilst attempting to achieve the all-important lowest efficacious dose effect; a consideration that is a rate limiting factor in the acceptance of a clinical trial application as exemplified by the Medicines Health Regulation Authority (MHRA) and by extension the European Medicines Agency (EMA) guidelines for good clinical practice in trials16. By extension when considering the construction of a clinical trial application, our data show that IRE1 or XBP1s would be a target worth investigating as a secondary objective biomarker17.

16 Such guidelines can be found in www.gov.uk/guidance/clinical-trials-for-medicines-apply-for- authorisation-in-the-uk and https://www.ema.europa.eu/en/human-regulatory/research- development/clinical-trials/clinical-trial-regulation 17 In oncology trials for aggressive tumours primary objectives are usually confined to survival only. Therefore biomarker discovery would come under the secondary objectives directive

236 The role of IRE1 in the maintenance of the GBM differentiated phenotype and in treatment resistance

When appraising a potential therapeutic target one should absolutely assess it from a target centric point of view to discern its importance in disease pathogenesis as carried out in chapter 4. However, a wealth of information may also lie in a response to current medicines perspective to determine what role it may play in the continuation of disease and the disease’s ability to respond to current therapies. After all GBM undergoes significant therapeutic regimes with surgery, radiotherapy and chemotherapy combining to what should amount to a devastating effect. However, the tumour almost always recurs. Considering IRE1’s involvement in the disease pathogenesis, what role could it be possibly playing in disease maintenance and GBM’s ability in evading chemotherapy? Answering this question would not only provide a more complete landscape of IRE1’s involvement in GBM benefiting the field of ER stress research but would also guide decisions as to the potential clinical applications of IRE1 signalling in GBM raising considerations of patient group stratification and treatment safety. In chapter 5 such considerations are addressed. It has been postulated that one way that GBM cells evade chemotherapy is by being able to de-differentiate back to a stem phenotype and hence evade the properties that chemotherapy targets. After all, chemotherapeutic agents target the archetypal cancer cell which is rapidly dividing, fast growing and uncontrollably proliferating. This is the reason for a lot of trademark signs of chemotherapy such as hair loss as it also targets other rapidly dividing cells in the body and alkylating agents such as TMZ bind onto DNA preventing its rapid replication. By reverting to a slow dividing, malleable, able to differentiate phenotype GBM cells may survive, adapt to and once the chemotherapeutic effect is washed out re-differentiate into more aggressive tumour cells that confer the known outcomes of GBM pathology. Rapidly dividing cells have high metabolic rates and we have already discussed in chapters 1 and 2 that highly metabolically active cells such as B cells and cancer cells display a massively expanded ER with increased UPR capacities to deal with the increased protein folding demand. It would thus stand to reason that IRE1 would be important in the maintenance of such a highly metabolically active phenotype.

Such a link had of course been suggested before but the specific underlying mechanisms had never before been as comprehensively analysed in as may experimental models. The cellular models described in chapter 4 proved valuable tools to the conduction of this investigation as both commonly available and primary lines were used in both cancer differentiated cell-like and cancer stem cell-like phenotypes to provide a detailed account of IRE1’s involvement in the differentiated state maintenance. In chapter 4 we uncovered the effect of RIDD on the state of miRNA levels and its involvement in GBM pathophysiology. Here we provide a definitive link

237 between a subset of miRNAs and XBP1s particularly focusing on miR148a and the transcription factor SOX2; players well described in their involvement in differentiation, reprogramming and malignant prognostication. It is immediately evident from our investigation that such a signalling axis is not exclusive and may be influenced by other factors. Indeed we provide evidence of other miRNA candidates (miR21) that could be under the influence of XBP1 signalling and should be considered for future investigation. However, as with the rest of the investigations in this thesis we have proceeded in an evidence based manner that dictated to follow the candidate (miR148a) indicated by the work of other experts in the field as the most relevant to our investigations based on its indicated involvement in GBM and the XBP1s signalling sphere of influence. Moreover, although SOX2 is the factor predominating our considerations we provide evidence for other transcription factors and gene sets that are directly or indirectly altered in by perturbing the IRE1-XBP1s axis, not only pointing to a gene cluster rather than monogenic effect of the UPR on differentiation but also to the effect that IRE1 targeting may have on different cell populations. After all each of the factors investigated in this study SOX2, SALL2, POU3F2, OLIG2, GFAP, TUJ1 can be traced as of exceptional interest in a particular neuronal cell lineage with involvement in neuronal embryonic development, ocular development, neuronal differentiation, oligodendrocytic lineage, astrocytic glial lineage and neuronal lineage respectively. None of these factors work alone, rather they are part of complex systems with the majority operating at least trimeric complexes which pertains to two considerations: 1) IRE1 signalling has exceptionally wide reaching consequences through its XBP1s arm that are not limited to the effects of XBP1s as a transcription factor signalling to the nucleus to induce genes related to protein control, 2) From a wider perspective it is likely that IRE1 is not the only factor influencing such complexes as cellular homeostasis is always built on redundancy and contingency; factors that have to be taken into consideration when scrutinising IRE1 as a therapeutic target as its blocking may be overridden by a different mechanism to restore selectivity towards differentiation and reprogramming and 3) IRE1 may play a limiting factor in its own effect as we have already described it can degrade miRNA through its RIDD signalling, adding to the potential of it regulating its own function in the cycle of de- differentiation and re-differentiation. It is worth considering that tumours that survive chemotherapy have a robust enough IRE1 signalling axis that initially would upregulate XBP1s splicing to facilitate miR148a induction and hence reprogramming transcription factor such as SOX2 downregulation succeeding in evading alkylating effects by reverting to a stem phenotype. Thereafter, once the shock of chemotherapy washes out, the RIDD pathway could take over to degrade the elevated levels of miR148a to allow for the increase of transcription factors such as SOX2 to allow for stem cells to re-differentiate and for the GBM tumour to recur.

238 These considerations aside, the investigation was carried out in a way that left little doubt as to the specificity of the IRE1 signalling axis in the maintenance of the GBM cell differentiated phenotype and the change to a stem phenotype in its absence. Multiple human commonly available and primary as well as mouse lines were utilised and the effects were observed by using both pharmacological and genetic manipulation of the IRE1 axis which included multiple different avenues of modulation including CRISPR KO, DN and Q* mutant overexpression, XBP1s overexpression and downregulation as well as miR148a manipulation by both mimics and inhibitors. This global approach has led to certain confidence as to the robustness of the investigation and its translational outcomes. Next steps should include investigating the other miRNA targets uncovered by the bioinformatic analysis as well as the literature analysis carried out at the beginning of this investigation. miR148a is indeed part of a trio of miRNAs alongside miR148b and miR152 that have been implicated in disease having similar roles due to their similarity in their recognition sequence. Throughout the investigation we saw the validation of the mechanism both genetically and pharmacologically. To keep up with the way this comprehensive investigation has been conducted, the effect of growing IRE1 deficient tumours in a GBM surgical mouse model should be investigated and finally the rate of tumour re-growth in the presence or absence of MKC8866 should be investigated to determine if indeed blocking XBP1s would cause faster recurrence.

What could these translational outcomes be though? As we outlined earlier in this discussion we are appraising IRE1 as a therapeutic target from the perspective of its effect on the response of GBM to current medicines. It was also stated that GBM cells evade chemotherapy by reverting to a stem phenotype. We have here postulated that XBP1s inhibition does the same thing. Does this mean that IRE1 should be discarded as a therapeutic option in GBM? It is here that multidisciplinary expertise lends its helping hand by providing us with clues from the clinic. Indeed using IRE1 inhibitors alone in specific GBM tumours could well have detrimental effects as tumours with high XBP1s activity could well revert to a stem phenotype and hence recur more quickly. Moreover, it is a well-established fact that IRE1 is a hugely important homeostatic mechanism that if repeatedly systemically perturbed (as could potentially be the case to achieve therapeutic effect in the tumour) would not only have off target effects but potentially decrease the quality of life of patients without having a significant effect on the tumour. However, the timeline of GBM recurrence, the timeline of GBM treatment and the timeline of de-differentiation has to be taken into consideration as the golden rule of timing being everything in medicine very much applies in this concept. A robust IRE1 signalling has to be in place in order for the tumour cell to ride out the insult conferred by TMZ. This treatment according to the current treatment protocol occurs over five post-operative days with

239 recurrence occurring over months. As such a truly attractive option would be to use IRE1 inhibition as an adjuvant therapy to TMZ given intra-operatively. Such administration would overcome the issues posed by the decreased delivery of drugs to the CNS due to the blood brain barrier and it would cripple the robust IRE1 response needed to adapt to TMZ. This way a larger proportion of terminally differentiated cells would perish and cells able to de-differentiate to a stem phenotype would not have the time to do so as they would be sensitised to TMZ and thus undergo cell death within the initial five days of treatment. The single intra-operative administration of IRE1 inhibition would also limit potential side-effects and the need for repeat prescription that could expose remaining cancer cells to persistent IRE1 inhibition that could trigger a reversion to a stem phenotype. As such IRE1 is a valid therapeutic target in GBM with the levels of XBP1/RIDD activity in each tumour a potential stratifying characteristic to determine IRE1 treatment in GBM.

IRE1 as a therapeutic target in GBM

At this point we should examine the advantages and disadvantages of modulating IRE1 activity as a therapy in oncology to put the above discussion into a wider clinical perspective whilst keeping in mind the results presented in both chapters 4 and 5. There are two wide aspects of IRE1 signalling that are taken into consideration when attempting to justify its use in cancer treatments and these are is propensities for both pro-survival and pro-death properties. On the one hand there is the notion of utilising an ER stressor as a therapeutic, where it would cause IRE1 hyper-activation in cancer cells to induce a terminal UPR and thus guide the cancer cell to apoptosis. One major barrier to this being a valid option is off course specificity and off target toxicity. As IRE1 and the UPR are ubiquitous throughout the human body, any such therapeutic would not only target cancer cells but also healthy ones. Considering that cancer cells are adapted to highly hostile environments and thus have a very robust UPR, the doses that would achieve an oncolytic effect would likely be extremely toxic to the surrounding tissue, which of course in the case of GBM is too precious to risk. As such over-activating the UPR may not be the solution.

What of blocking it? We have established that tumour cells have a robust UPR and that any therapy designed to over-induce it would be too toxic for the other cells to cope with. However, stressing the UPR is not only achievable through ER stressors. After all most chemotherapies are toxic enough to induce ER stress by themselves. The UPR as we covered in the introduction of this thesis is a fine-tuned balancing act between PERK, IRE1 and ATF6 with PERK involved in the pro-death pathway through CHOP. It is thus not extraordinary to consider

240 that blocking one of three UPR transducers would derail this balance overburdening the other two arms of the UPR potentially pushing PERK towards signalling apoptosis and hence giving IRE1 inhibition some potential value as a monotherapy in cancer. Moreover, it has been postulated that PERK is involved more towards the end-stages of the disease with IRE1 more involved in tumour development, a state of affairs that could be exploited along with strict diagnostic and prognostic timelines to guide the potential of IRE1 inhibition as a treatment option. Once again, the dose escalation for such an operation would be complex and painstaking with uncertain outcomes but as we established in chapter 4, not all patients display the same levels of IRE1 activity. It would thus be prudent to prescribe an RNase inhibitor not to patients with tumours of low IRE1 activity as this could be a sign of cancers that are adapted to work without IRE1. As such treating tumours displaying high IRE1 activity with IRE1 inhibitors would potentially block a pathway that those GBM cells rely upon and hence weaken them to the hostile tumour microenvironment of hypoxia and low pH that solid tumours have to survive. Further to this we showed in chapter 4 that high IRE1 activity tumours had a worse prognosis than low IRE1 activity tumours. It is thus conceivable that by treating with an IRE1 inhibitor we could force tumours into the “low IRE1 activity” category, thus using it as a disease modifying treatment to improve prognosis, as indeed a small increase in survival could have profound value to a patient, their family and their support system. In this capacity IRE1 could also be used as a disease modifying treatment aimed at slowing progression through its potential capacity as an antiangiogenic, since most studies that have investigated the involvement of IRE1 in GBM have explored its effect on blood vessel formation.

A major consideration against this line of thought of IRE1 inhibition as a monogenic therapy is that currently the most potent IRE1 inhibitors are targeting the RNase domain of IRE1 thus completely blocking both XBP1s and RIDD signalling which as described in chapter 4 not only may have opposing biological outputs on angiogenesis and invasion but also have different prognostic outcomes and as described in chapter 5 the XBP1s arm may favour a stem phenotype that can evade chemotherapy. As such although completely blocking the RNase domain may be the right option for a subset of tumours, there are enough warning signs to not be considering this as an option for all GBM patients. Another characteristic of IRE1 RNase inhibitors is that they are reversible and hence would need prolonged doses in order to achieve a long lasting effect something that would not work in the favour of patients with tumours with a robust UPR response (which would survive the inhibition) that could utilise such XBP1s blockade to revert to a stem phenotype.

It therefore seems that the landscape of IRE1’s attractiveness as a therapeutic target is dependent on a multitude of factors that should form an environment that would over-stress

241 tumour cells to induce apoptosis, whilst at the same time limiting their ability to adapt by inhibiting IRE1 in a strict time frame, without completely blocking its RNase activity depending on the type of GBM tumour treated. It is a daunting task however, combinatorial therapy of surgery, chemotherapy and radiotherapy (ER stress inducers alongside DNA damagers) and a short term intra-operative dose of IRE1 inhibition to sensitise GBM cells to an onslaught of therapeutic measures would fulfil the majority of the above criteria whilst limiting the chance for off target effects associated with prolonged exposure and thus would most likely be the best option for utilising IRE1 inhibition in GBM.

Discovery of a potential IRE1-derived and IRE1-targeting, small molecule adjuvant treatment alongside TMZ against GBM

The next logical step was to test this sensitisation theory in GBM whilst pursuing the development of a robust, evidence based drug discovery pipeline to design specific IRE1 modulators. More considerations had to be taken into account though. As we discussed in the introduction and subsequent chapters, specificity would be a limiting factor in IRE1 inhibition. Moreover, accessibility of any potential novel compounds to the clinic would be of importance when assessing their advantage over existing modulators or repurposing existing clinically approved medications. To this end the IRE1 structure itself was exploited since fragments of its luminal domain had already been shown to affect its activity in vitro and in vivo, it being a substrate of itself. After all, what better way to ensure specific binding and homology between receptor and ligand than deriving the latter from the former? As such a library of small peptides derived from the IRE1 luminal domain was generated and their binding properties to the IRE1 kinase domain were identified in silico. The kinase domain was chosen for a specific reason over the ribonuclease domain. Firstly, exceptional ribonuclease inhibitors already exist (MKC8866) that specifically and robustly block the IRE1 RNase activity. Secondly we have established that the IRE1 RNase output is 1) beneficial to normal physiology, thus raising the possibility that blocking it completely may induce unwanted side effects and 2) has a dual effect in GBM. We have established that XBP1s promotes macrophage infiltration, angiogenesis and invasion whilst RIDD attenuates such effects in GBM. Moreover we have established that XBP1s rather than RIDD plays a role in the de-differentiation of GBM cells. As such completely blocking the activity of two at times antagonistic signalling pathways would not be of overall benefit. Contrastingly maintaining some IRE1 activity whilst targeting its kinase domain may induce conformational changes that affect its trans-auto-phosphorylation and oligomerisation thus dampening its activity rather than killing it altogether.

242 Picking small peptides that fit the binding pocket that ATP/ADP would normally occupy with better binding properties than ADP/ATP we ensured that the amino acid homology would confer some sort of selectivity for IRE1 binding since binding on every single kinase pocket in a cell could have off target effects. Using the binding properties of these peptides we came up with four clinically and two non-clinically approved compounds that had better binding capabilities than ATP/ADP on the IRE1 kinase domain pocket in silico. These compounds along with the peptides that were used for their discovery were confirmed to have a negative effect on the IRE1 RNase activity in vitro. RNase activity was used as the outcome and an RNase inhibitor (MKC8866) was used as the control because the functional outcome of IRE1 is through its ribonuclease rather than its kinase activity. As such, the information was provided to be able to proceed to more complex cellular models of GBM to determine if a similar effect was seen in these. Interestingly in vitro, certain compounds such as folinic acid and cefoperazone were shown to be better inhibitors of RNase function than others such as methotrexate and the Z compounds. A structural approach could potentially explain such an effect as folinic acid and cefoperazone seem to share some structural similarities in their backbone as MKC8866. As such it could be that these compounds are able to bind both the kinase and the RNase domains of IRE1. This is a particularly interesting concept in the case of folinic acid and methotrexate. These two are co-prescribed in both immune diseases and cancer with folinic acid used as a molecule to decrease the side effects conferred by methotrexate. In this context the toxicity of methotrexate and folinic acid could well be attributed to the targeting of IRE1 and their inter- limiting capabilities due to their variable effects on IRE1 function.

As kinase inhibitors/ ATP competitors all six molecules seemed to have an effect on IRE1 phosphorylation offering some confirmation of their mechanism of action and justifying the in silico approach that preceded them. Moreover they had a marked effect on the levels of XBP1s in both commonly available and primary GBM cells in the presence of two different ER stressors, pertaining to a lack of ER stress mechanism specific IRE1 inhibition; a desirable trait as the ER stress that GBM tumour cells undergo is due to a multitude of factors adding to the translatability of these observations.

Having shown that designing a drug discovery process by which already existing molecules that can inhibit IRE1 activity in silico, in vitro and in cell-based models can be identified, we have observed the development of an indispensable tool to by-pass novel drug synthesis and aid drug repurposing, accelerating their testing in new clinical settings. Through this tool novel modes of action for molecules already in abundant clinical use were identified as well as molecules with no clinical designation. A new potential clinical designation was

243 presented for these newly described IRE1 inhibitors as we showed that they sensitised both commonly available and primary GBM lines to TMZ.

We structurally interrogated the IRE1 kinase catalytic domain and subsequently carried out an in silico screen to have identify four clinically approved molecules and two novel, hitherto not clinically approved molecules, that inhibit IRE1 RNase activity in vitro and in human GBM cell models, and sensitise them to the established GBM chemotherapy treatment, TMZ. As discussed in chapter 3 of this thesis such discoveries are of great benefit to clinical trial design as molecules can be put straight into the clinic with much reduced temporal and financial burden onto the design of the trial itself. This means that patients would benefit from such insights in an expedient manner.

It is of course prudent to discuss here the next steps of such an investigation. Despite the fact a humanised cell model has been utilised to great effect, it cannot provide information on a physiological or anatomical level. This investigation would thus greatly benefit from an animal surgical model of disease to determine if the molecules identified have an effect on tumour growth and overall survival as well as carry out preliminary investigations of toxicity at an organismal level. Apart from investigating at the higher organism level, it would be beneficial to explore the other end of the spectrum and explore avenues of research to verify the exact mechanism of action of such molecules on IRE1. We have postulated the possibility of dual RNase/ Kinase activity or even the possibility of a truly allosteric inhibitor that would bind neither the kinase nor RNase pocket. Such questions would benefit from further in silico work combining molecular dynamics (MD) simulations and binding hypotheses to assess the binding of such molecules on other regions of IRE1 or indeed the generation of allosteric inhibitors that would address the concern of off target binding due to kinase pocket structure conservation throughout many kinase molecules. The MD simulations would aid in determining what quaternary structure IRE1 assumes when bound to these molecules. As any conformational change is key to its function such investigation would shed light onto the mechanism of inhibition. As such oligomerisation assays would also add to the wealth of information gathered as they would provide clues as to whether kinase inhibition and ATP competition by these compounds affects IRE1 oligomerisation and hence downstream cleavage of XBP1 and mRNAs/miRNAs through RIDD. Finally, the generation of a drug resistant IRE1 mutant and its expression in GBM primary lines would provide a control to verify that the sensitisation to TMZ observed in the cell lines is due to IRE1 inhibition.

244 Perspective: Bringing IRE1 as a therapeutic target to the patient

We have established through a substantial body of work that IRE1 is a valid therapeutic target in GBM and we have even overseen the identification of a series of inhibitors that would sensitise GBM to TMZ. At this point (and of course throughout this investigation as evident in the results and the discussion chapters) reflection is of great importance. We have scientifically justified the potential of inhibiting IRE1 in GBM. The next step is infinitely more complex as we are proposing an interventional medicinal product for a disease with dire consequences. As such our questions firmly shifts from why and how, to who is going to receive this treatment.

With such a question comes great responsibility and have to therefore carefully consider the alternative outcomes beyond initial deliberations of safety and toxicity. Who would this treatment concept benefit most? As we saw in chapter 3, GBM is a vastly heterogeneous disease with a lot of tumours displaying intra-tumoral plasticity. In addition the molecular phenotypic range of mesenchymal, pro-neural and classical is enriched by the prognostic and diagnostic genetic demarcation of the presence or absence of IDH mutations, MGMT promoter methylation or mutations on EGFR, TERT, PTEN, TP53 and others. It thus, logically, follows that such heterogeneity would apply to the range of responses expected to any potential novel IRE1 targeted therapy. An already complex landscape would benefit from further information and we have attempted to do just this but from a functional perspective. By assessing the status of IRE1 in patient tumours we were able to have an idea of how tumours would react to their microenvironment and the subsequent treatments rather than only taking into account the prognostication afforded by established genetic signatures. The robustness of such IRE1 dependent classification is of course by no means concrete and it should be investigated further as the transcriptomic data that it represents are showing but a small time frame in the life of a tumour, a rather difficult obstacle to overcome since repeat biopsies would require repeat brain surgeries, operations with associated comorbidities and of taxing nature to the patient. XBP1s levels in the CSF could be a potential guide as to the tumour propensity towards IRE1 activity, although even then such a read would not be definitive as a patient with a brain tumour would have higher levels of ER stress markers than normal.

Regardless, our accumulated information could potentially shed some light onto the opportune moments of IRE1 inhibition in GBM. As no single definitive marker for RIDD activity is as of yet identified, by utilising XBP1s one could stratify patients further. It would not be unimaginable to propose a treatment plan for a patient with an IDH mutant, MGMT promoter methylated, XBP1s high activity tumour that would include IRE1 inhibition as single dose adjuvant therapy along with TMZ and radiotherapy as a patient would benefit from a tumour

245 with better prognosis (IDH mutation), that could cope less with TMZ induced DNA alkylation (MGMT methylation) and would not be able to adapt to the overall stress induction due to the lack of a fully functioning UPR through IRE1 blockade (Low IRE1 function-better prognosis). By the same token since we have established that RIDD may play a role opposing angiogenesis and invasion, tumours with high RIDD activity would potentially benefit from antiangiogenic treatment but not IRE1 inhibition. Since allosteric IRE1 inhibition may affect RIDD activity more than XBP1s activity by blocking high order oligomer formation one could utilise such inhibitors in the presence of chemotherapy to sensitise GBM cells to the treatment without potentially inducing a shift towards the stem phenotype, an issue that could itself be limited by the temporal aspect of treatment planning.

As such, pharmacovigilance is of great importance. Targeting a homeostatic mechanism is accompanied by great risk, but with robust patient stratification and educated and careful good clinical practice the benefits to the patient could be life enhancing. Advice by a medicines health regulation authority as to the optimal steps necessary to address such considerations after addressing all pre-clinical questions would have tremendous added value to the effort of bringing IRE1 inhibition to the clinic. We briefly touched on in chapter 6 and earlier in this discussion the need for an animal model to accumulate pre-clinical data on the physiological effects of a potential IRE1 inhibitor in GBM. However, such studies form the backbone of pre- clinical drug toxicity profiling as well, whilst observing principles of humane experimentation in reduction (decrease in numbers of animals used to obtain information of a given standard), refinement (decrease in the incidence and severity of procedures applied to animals) and replacement (substitution of conscious living vertebrates by non-sentient material). When interpreting such efficacy and toxicity tests there are a number of factors to consider such as reproducibility, consistency and biological plausibility, inter-species heterogeneity, metabolism and changes in kinetics/dynamics, the number of animals used for observation, the reversibility of any effects observed and last but not least the relevance of effect to humans. Such considerations fall under the remit of Good Laboratory Practice (GLP) which is an essential part of quality control of pre-clinical data that are assessed for clinical use to support reproducibility and reliability.

GLP, although strictly enforced in pre-clinical studies, is very much a precursor and partner of the term GMP or Good Manufacturing Practice which describes a process capable of consistently generating products, through validated critical manufacturing steps, in appropriate facilities using suitable equipment by qualified personnel. In our case of intra-operative IRE1 inhibition in GBM this would not only apply in the production of a medicinal compound but also to the method of delivery, briefly alluded to in chapters 5 and 6. These processes would

246 enable quality assurance, an umbrella term for both GMB and Quality Control, for the generation of a medicinal product and a medical device in the form of a sponge or plug that would be placed in the resected tumour cavity to ensure direct but slow releasing administration of IRE1 inhibition by at the same time bypassing obstacles such as the blood brain barrier. Such a device since implantable would be classed as a high risk device and would thus need to be approved and CE marked through a clinical investigation potentially in tandem with the medicinal product it would be delivering (in our case an IRE1 activity modulator). The interconnected nature of the concepts of GLP, GMP, pre-clinical and clinical development of device and a drug can be seen Figure 1. “Incorporation of GLP and GMP in pre-clinical and clinical development of an intra-operative method of IRE1 inhibition administration in in figure 1 of this discussion. GBM”. Conceptual design of the workflow of development of an animal surgical model of GBM, a device as a method of intra-operative administration of IRE1 An IRE1 inhibitor inhibition and IRE1 inhibition as a therapeutic. This flow encompasses good laboratory and manufacturing practice throughout pre-clinical and clinical would enter the clinic development, concepts important not only in the process of bringing IRE1 inhibition to the clinic and market but also to the clinical trial application through a series of clinical process itself. trials termed CTIMPs or

247 clinical trials of investigational medicinal products which would differ to clinical trials designed to test devices or food products. Such trials have notoriously high attrition rates and high costs with only 5% of novel therapeutics entering a Phase I clinical trial in oncology reaching the market. This ensures enormous pressure on the design of such a trial as the future of the therapeutic intervention depends on it with both clinical and financial implications. As such the consideration of how an IRE1 inhibitor would enter such a setting is of importance when pondering the future steps of bringing such a treatment to the clinic. Usually, CITMPS are starting with what are called sentinel trials were new medications are tested for toxicity in single healthy volunteers before entering Phase I testing were traditionally 6-12 healthy volunteers are tested for toxicity. There are two major differences between such trials and equivalent trials in oncology. Firstly healthy volunteers are not used at all with cancer patients taking part instead. Secondly, a Phase I clinical trial for any other CTIMP would prioritise dose scheduling and toxicity; in oncology efficacy is an observational outcome as well. The reasons for this dichotomy are obvious. For patients that are faced with terminal diagnoses of aggressive tumours or for patients who have not benefited from any of the already available medications or procedures taking part in a clinical trial with potential prognostic advantages supersedes the potential toxic side-effects. This is particularly of note in oncology trials as placebo treatments are never used so patients participating are always receiving the medicinal product under investigation.

There are a few Phase I clinical trial designs including accelerated titration, continual reassessment and pharmacological guidance system designs. However, the most commonly used is the three by three trial design, where novel therapeutics are tested in pools of patients with aggressive tumours of varying origins such as breast, prostate, liver or brain. In such a set- up, 3 patients would receive a given minimal concentration of the interventional therapeutic in question (in our case IRE1 inhibitor). The next 3 patients would receive a higher dose and so on until adverse effects are observed producing the maximum tolerated dose. Using this as a benchmark of toxicity, the rest of the trial (Phase 1b) may be carried out where the study treatment can be tested against the current treatment of choice or as an adjuvant therapy alongside it. There is no intra-patient dose escalation and these toxicity findings adhere to grading scales such as the NCI common toxicity criteria.

248 Considering bringing IRE1 to the clinic we may envisage combining characteristics of clinical trial design to bridge the findings presented through chapters 4-6. An additional pre- Phase I step that would benefit a thorough investigation of IRE1 inhibition in GBM is a phase 0 study where imaging such as PET scanning is used to determine whether the drug is reaching the tumour. As such, a radioactively labelled molecule would be tracked to the tumour site enhancing its validity as an investigational therapeutic. This would also provide information as to the absorption in surrounding healthy tissue. As established earlier, patient stratification based on IRE1 or XBP1/RIDD activity would be of great interest to guide the study population. In order to determine the IRE1 identity of each tumour one would have to avoid major surgery reserved for tumour excision but also obtain readings of IRE1 activity rapidly. To this end stereotactic needle biopsy could be combined with next generation sequencing. The reason for circumventing major surgery is that brain operations carry considerable risk and associated comorbidities, hence should be reserved for the excision of the tumour where possible. In this case, if Figure 2. “Phase I trial of an IRE1 inhibitor in GBM”. Conceptual design of an oncology GBM trial encompassing points made throughout the discussion based considering administering an on findings in chapters 4-6. The choice of group 2 in this instance is hypothetical rather than suggestive as to what signature would mostly benefit from IRE1 IRE1 inhibitor inhibition, since other tumour characteristics have to be taken into consideration. Elements for the schematics in this figure have been adapted intraoperatively, we would nd from Netter's Clinical Anatomy 2 Edition, 2010, Saunders Elsevier. Brain need to know the IRE1 sections in red represent patients who display toxicity side-effects. identity of the tumour before

249 the surgery itself and thus would justify the need for a biopsy. In order to maximise the sample size of the study, it should be multicentre as well as international (where possible) to cover a maximal range of demographics and to ensure that enough patients would be recruited to participate in the study for clinically meaningful results. A schematic of what a phase I clinical trial of IRE1 inhibition in GBM would look like can be seen in figure 2 of this discussion.

Once the IRE1 signature for each tumour has been identified the established treatment protocol of surgery, chemotherapy and radiotherapy can be followed with the addition of escalating doses of IRE1 inhibition to determine toxicity using the 3 by 3 trial design. Due to the targeted nature of the investigation, the investigational therapeutic would be tested on GBM patients only. Consequently, an intra-trial stratification design would permit running parallel toxicity studies were groups of patients with different IRE1 signatures are treated with IRE1 inhibitors as adjuvant therapy to TMZ evaluating the effect that IRE1 inhibition may have on XBP1 high or RIDD high tumours. Therefore there would be multiple intra-study observational and therapeutic objectives that would support a targeted approach to exploiting IRE1 in GBM, especially due to the high attrition rates observed in oncology trials in general, making its admittance in the clinic more expedient provided the outcomes of the Phase I and subsequent studies would be as predicted. Since patient stratification based on IRE1 activity is a major outcome of the investigations described in chapters 4-5 and a major informative advantage in designing a study that not only assesses efficacy and toxicity of IRE1 inhibition but also a prognostic and treatment predictive biomarker signature based on IRE1 activity; patients recruited would have to give fully informed consent to their subjection to IRE1 treatment. Understanding of the risks associated with such treatment would be of high priority to the physicians obtaining consent but also studies as those described earlier in this thesis would be part of the rationale of the trial design and hence part of the cost-benefit analysis that patients and physicians would have to go through in order to determine if a patient diagnosed with a GBM tumour would be eligible to participate.

As such, although further pre-clinical data would need to be accumulated to support the conveyance of IRE1 inhibition to GBM therapy there is a rational path for such molecules to be tested in clinical trials for efficacy and toxicity in groups of GBM patients clustered according to the IRE1 identity of their tumour through an oncology, multicentre, randomised Phase I clinical trial.

250 Conclusions

We have shown that IRE1 may play a differential role in GBM pathophysiology through angiogenesis and growth as well as adaptation to chemotherapy and maintenance of differentiated GBM cell phentoype. These effects are conferred through XBP1s and RIDD signalling. Collectively we have shown XBP1s signalling to promote macrophage infiltration to the tumour, angiogenesis, invasion and maintenance of a differentiated aggressive phenotype, whilst RIDD may attenuate angiogenetic and invasive properties as well as possibly keep the miRNA environment in check playing a role in controlling cell re-differentiation in GBM. As such we have provided ample evidence of the importance of IRE1 in GBM pathophysiology and tested its validity as a therapeutic target by generating translational cellular models of GBM, introducing genetic IRE1 modulation in these models and testing their sensitisation to TMZ in the presence of targeted IRE1 kinase inhibitors by establishing a novel drug discovery pipeline and producing six as yet unknown to impact IRE1 activity inhibitors.

As such this author has:

1. Generated a translational cellular model of altered IRE1 activity in GBM. 2. Characterised the impact of IRE1 signalling on GBM pathophysiology and treatment resistance showing its effect on tumour growth, angiogenesis, invasion and tumour stem cell reprogramming. 3. Generated novel IRE1 modulators and tested their effect on GBM sensitivity to temozolomide; the current chemotherapeutic option in GBM, showing that targeting IRE1 is indeed a valid pharmacological target in GBM cell sensitisation to chemotherapy.

Therefore IRE1 is a major pathophysiological mediator and valid pharmacological target in GBM and its modulation may provide novel therapeutic options as an adjuvant disease modifying treatment.

251 Epilogue

ǲ Ǥǡ ǡǡ ǡ Ǥǳ

Marie Sklodowska Curie, Autobiographical Notes

I have many reasons to be grateful to Marie Sklodowska Curie. The actions in her name have provided me with a unique opportunity as an ITN Fellow to maximise the potential of positive impact my translational medical research can have on the public and my personal development. Now, her autobiographical notes perfectly illustrate the culmination of three years of work as only by improving can our work in the laboratory have a chance to make a difference in the clinic. Having experienced both these worlds of science and medicine, it is evident that the bridge between them, (that is of greatest benefit to patients) can only be built upon strong exchange of excellent ideas. I have been lucky to be able to discuss, collaborate and exchange across countries, cultures and disciplines to improve myself and provide work that might, just might, be of benefit to someone who will need it.

It has been hugely inspiring and immensely productive and thus an experience that I will never forget.

252 Special Thanks

First and foremost to Dr Eric Chevet for the opportunity to participate in such exciting projects and for 3 years of positivity and motivation, culminating in a working environment of smooth collaboration, success and constant communication! My time in Rennes absolutely flew by, denoting how much it was relished and the results speak for themselves. Neither the volume of work described in this thesis nor the short timeline within which it was achieved would have been possible without Eric’s input and unrelenting responsiveness and positive attitude. As such, I do not think I could have asked for a better supervisor.

To Dr Tony Avril for all the support and help to realise some of the major goals of this project and as well as the constant willingness to communicate and make time alongside his numerous other projects.

To Prof. Leif Eriksson (Gothenburg, Sweden) who provided me with an opportunity to do something completely new to me and to work in yet another different environment, whilst providing constant support and communication during my time in Sweden and thereafter.

To Gwenaele Jegou who helped me find my feet in the lab and provided ample training when I first arrived, to Celine Lepine and Jeanne Bablee, without whom any of the indispensable administrative work could have been achieved and to Raphael Pineau and Florence Jouan who undertook significant work on the in vivo models.

To Prof. Jackie de Belleroche (Imperial College, UK) who took an active interest in my future and helped shape the qualifications and scientific background necessary for me to acquire this position

To Alexandra Papaioannou and Joanna Obacz who were present from the start to the finish of these last three years and shared the experiences of this period with me in full.

253 To everyone in Centre Eugene Marquis who made me feel welcome and made the last three years truly enjoyable. Coming to a new country is never easy and to be able to hit the ground running and produce as much in as little time speaks volumes of the productive atmosphere created by everyone in the COSS team in particular and at Centre Eugene Marquis generally.

Alice Metais Sophie Martin

Pierre Jean Le Reste Jerome Archambeau

Kim Barroso Laura Chaillot

Mari McMahon Chloe Sauzay

Héloise Bourrien Alice Blondel

Melissa Thomas Daria Sicari

Emna Mahfoudi Celia Leon

Vesna Risso Diana Pelizzari

Matthieu Le Gallo Xingchen Zhou Elodie Lafont

To everyone in the TRAIN-ERS network who made all our international meetings a fantastic occasion, Dr Sandra Healy in particular as the chief organiser of these meetings. It has been excellent to be part of such a multicultural and multidisciplinary group and share a lot of fun times and professional experiences, whilst being able to visit a lot of different places across Europe.

Aitor Almanza Luigi Montibeller

Chetan Chintha Sanket More

Stuart Creedigan Franzi Pueschel

Brian Leuzzi Maria Sassano

Andreia Luis Joseph Skoko

Nicole McCarthy To everyone in the Eriksson lab and Antonio Carlesso in particular, in Gothenburg for providing a fantastic opportunity to collaborate on a completely novel scientific field for me. They provided a fantastic atmosphere to maximise output whilst enjoying life in Sweden as much as possible over three thoroughly unforgettable months.

Anna Reymer Jae Ho Shin

Sonali Chavan Sergi Gomez Ganau

Chunxia Gao Samuel Genheden

Johanna Hoerberg

To everyone in eNIOs Athens who provided invaluable input either in person or from afar. Without their input a lot of the work presented in this thesis would not have been possible

Aristotelis Chaziioannou Efstathios Vlachavas

Konstantinos Voutetakis

254 To everyone in CMI Rennes and Ecole Doctorale who helped me along the way with everything important outside work including finding a place to live, dealing with administrative issues and arranging all my liaisons with the University.

Luc Manoeuvrier Badia Msassi

Emilie Gesnys

To the members of this thesis committee who dedicated their time and effort to go through this manuscript, provide me with exceptionally encouraging comments as to the quality of the work and be present on the actual day of my thesis defence.

Dr Fabienne Foufelle Prof. Sophie Janssens

Dr Stéphane Rocchi Prof. Nicolas Pallet

Last but not least to all the people that I met throughout these past few years in Rennes that were not part of my professional environment. They made life very enjoyable indeed and helped paint an idyllic picture of Rennes that I’ll carry with me for a long time.

255 Résumé de thèse

Ce résumé présente, chapitre par chapitre, le contenu de la thèse de doctorat intitulée «Cibler l’activité IRE1 dans le glioblastome multiforme»; écrit par D. Doultsinos

Le premier chapitre de cette thèse a un double objectif. Premièrement, il nous donne l’opportunité d’avoir une vision globale de l’ER comme organite et donc d’examiner l’effet papillon potentiel2 de sa perturbation sur la physiologie normale et par conséquent sur de nombreuses pathologies. Le réticulum endoplasmique (RE) est le site de production et de repliement d'environ un tiers de toutes les protéines cellulaires; une contribution stupéfiante à la vie cellulaire et à l'organisme. Son fonctionnement harmonieux est d’une importance capitale pour la vie des organismes allant des archées à l’humain. Cette fonctionnalité oscille autour d'un état de repos susceptible de changer, et donc de stresser, à travers une multitude de facteurs, qui peuvent être aussi bénins qu'un changement de température saisonnier ou la division d'une cellule au cours du développement. Des voies de signalisation spécifiques, telles que la réponse protéique non pliée (UPR), sont responsables de la réponse et du contrôle de l'amplitude de ces oscillations. La qualité de leur réponse détermine le destin de la cellule, ce qui place l’urgence et l’EPU au centre de l’équilibre entre santé et maladie; vie et mort. En tant que tel, l’impact des urgences sur d’autres organites et le comportement cellulaire et son exploitation potentielle pour répondre à des besoins biomédicaux non satisfaits est d’une valeur évidente.

Deuxièmement, il s'agit de l'examen et du résultat d'un exercice d'écriture collaborative en tant que cible réalisable d'un groupe de scientifiques débutants couvrant une variété de sujets centrés sur les ER dans 14 projets de recherche menés dans 8 centres de recherche européens dans le cadre d'un réseau de formation innovant " «TRAIN-ERS», financé par la Commission européenne, dans le cadre des actions Marie Sklodowska Curie. En tant que tels, nous pourrons examiner les avantages de la collaboration interdisciplinaire et les résultats ultérieurs d'une telle entreprise par le biais d'une analyse documentaire approfondie.

Dans ce premier chapitre, nous avons documenté les connaissances sur les urgences qui ont conduit à ses rôles émergents dans la physiologie et la maladie, englobant les possibilités de communication et d’exploitation interorganes dans le domaine de la bioingénierie et de la santé. des outils pour guider l’innovation thérapeutique et la stratification dans des bassins de patients spécifiques. Nous avons étudié la manière dont la signalisation conservée dans les métazoaires permettait d’étudier la signalisation du stress dans différents organismes. Par conséquent, la recherche sur divers organismes est complétée par des stratégies à haut débit et in silico permettant de produire des modèles in vitro et in vivo. nous avec des données pour l'examen pré-clinique et l'application clinique ultérieure. Il est donc évident que la compréhension de la signalisation ER et de la réponse à l'EPU peut améliorer les soins de santé aux niveaux préventif, diagnostique et thérapeutique. Dans le chapitre suivant, nous nous concentrerons davantage sur l'EPU et l'IRE1 et sur les moyens de les cibler.

Au chapitre 1, nous avons exploré le rôle du RE dans la physiologie normale et avons approfondi les minuties de sa signalisation du point de vue de la RUP. Nous avons établi comment l'UPR agit en tant que médiateur majeur de la réponse des urgences au stress physiologique et avons décrit comment la signalisation UPR peut être exploitée et est actuellement exploitée dans des applications cliniques. Nous avons parlé de l'utilisation de techniques in silico et à haut débit nous permettant de construire des cibles pharmacologiques et de permettre à la découverte de biomarqueurs d'amener les molécules de ciblage ER et UPR du banc au lit, en examinant des exemples particuliers d'essais cliniques centrés sur les ER et des listes de modulateurs UPR. Dans le chapitre 2, nous rapprocherons les concepts de signalisation UPR, d’identification de molécules et d’application de maladies en nous concentrant sur l’UPR, en utilisant des scénarios cliniques spécifiques pour décrire ses diverses décisions relatives au destin cellulaire. En tant que tels, nous analyserons ses propriétés favorables à la mort à travers l'exemple de la sclérose latérale amyotrophique (SLA) et nous examinerons d'abord la relation entre le glioblastome multiforme (GBM) et l'IRE1 lors de la description de ses propriétés favorables à la survie. Ensuite, nous résumerons les méthodes utilisées par les rapporteurs cellulaires de l'activation de l'EPU pour identifier les modulateurs de l'EPU. Nous examinerons en outre les rapporteurs de la fonction ER afin de déterminer les efforts déployés pour identifier les molécules ciblant la voie de la sécrétion. Pour achever la revue des processus utilisés dans la découverte de modulateurs, très sélectifs pour l'EPU, nous aborderons les stratégies in silico qui traitent des problèmes de découverte ciblée des modulateurs d'EPU tels que les mécanismes compensatoires contre la modulation et l'existence de paralogues hautement homologues des cibles. Ensuite, nous nous concentrerons davantage sur l’activité IRE1 et discuterons des molécules qui la ciblent et de leur utilisation dans des modèles de maladie. À la fin du chapitre 2, nous aurons exploré le rôle de l'EPU, et de l'IRE1 en particulier, dans le traitement des maladies et discuté des processus connus en place qui facilitent l'identification et l'exploitation de modulateurs spécifiques à l'EPU. Dans ce deuxième chapitre, nous avons examiné la signalisation canonique et non canonique de l’EPU en consolidant les informations acquises au chapitre 1. De plus, nous avons revisité les principaux acteurs de la signalisation UPR, PERK, ATF6 et IRE1, et nous avons commencé à nous concentrer sur IRE1. Le médiateur de l'EPU a conservé l'évolution dans son évolution, devenant ainsi la pièce maîtresse de cette thèse. Dans une perspective large des stratégies de dépistage disponibles et des rapporteurs de la fonction ER et de l'activation de l'UPR, nous avons résumé et analysé les approches in vitro de la découverte et de la synthèse du modulateur UPR. Poursuivant notre cheminement des fonctions générales des urgences à l’examen périodique universel, de la découverte de médicaments à la formulation de l’hypothèse de cette thèse; IRE1 est désignée comme cible pharmacologique dans divers modèles de maladies et nous avons brièvement exploré son association avec la physiopathologie des GBM.

Au chapitre 1, nous avons exploré le rôle du RE dans la physiologie normale et avons approfondi les minuties de sa signalisation du point de vue de la RUP. Nous avons établi comment l'UPR agit en tant que médiateur majeur de la réponse des urgences au stress physiologique et avons décrit comment la signalisation UPR peut être exploitée et est actuellement exploitée dans des applications cliniques. Dans le chapitre 2, nous nous sommes concentrés davantage sur le rôle de l'EPU dans la santé et la maladie et nous avons commencé notre travail sur la signalisation et le ciblage de l'IRE1 dans divers scénarios cliniques, notamment le principal objectif clinique de cette thèse, le glioblastome multiforme (GBM). Au chapitre 3, nous présenterons la GBM en tant qu’entité clinique. En utilisant, dans un premier temps, une approche chronologique, l’histoire de la GBM sera décrite à partir de ses premiers rapports connus jusqu’à sa classification et compréhension actuelles. Après avoir traversé cette phase initiale qui démontrera les avancées de connaissances acquises au cours des 200 dernières années et soulignera les problèmes liés à la complexité de la caractérisation d’une maladie aussi hétérogène que la GBM, nous approfondirons les caractéristiques cliniques de la maladie. Partant d'une perspective de santé publique analysant sa prévalence et son incidence selon son pronostic établi, le fardeau de la GBM sur la médecine de population sera analysé. Ensuite, nous expliquerons les options cliniques disponibles pour les patients atteints de GBM et comprendrons les obstacles inhérents à une thérapie réussie dans le traitement d'une tumeur cérébrale invasive, tout en donnant un aperçu du paysage thérapeutique actuellement développé pour être testé dans la GBM. Ayant pris en compte l’état actuel du développement thérapeutique dans la GBM, nous allons encore combiner la signalisation IRE1 et la physiopathologie de la GBM, en fournissant un lien de continuité avec le chapitre 2 et en expliquant pourquoi le ciblage de IRE1 dans la GBM peut être une option thérapeutique valable basée sur ce qui a déjà été décrit aux chapitres 1 et 2 concernant l'importance de l'IRE1 dans la physiologie et la maladie et le rôle que l'IRE1 a été décrit jusqu'à présent dans la physiopathologie de la GBM. Dans ce troisième et dernier chapitre de l'introduction, nous avons passé en revue le concept clinique de GBM et avons expliqué pourquoi il méritait l'attention qu'il mérite de la part de la communauté scientifique clinique et translationnelle en tant qu'entité clinique. C'est une maladie lamentable dont il reste encore beaucoup à découvrir quant à ses origines et à son évolution. Après avoir exposé l'historique qui a façonné la classification moderne de la GBM, les options thérapeutiques actuelles pour les patients ont été examinées. Bien qu'il soit évident que la recherche sur la GBM bénéficie de toutes les techniques et méthodes de recherche novatrices disponibles, elle nécessite encore une caractérisation plus poussée aux niveaux moléculaire et macroscopique afin de parvenir à une véritable stratification du patient et, partant, d'améliorer les résultats du traitement. Après avoir résumé les essais cliniques prometteurs en cours sur la GBM, en se concentrant sur les essais de phase III, il est apparu que des thèmes tels que la souillure, la différenciation, la réutilisation de médicaments, le traitement à base de petites molécules, la sensibilisation de la tumeur aux traitements actuels par le biais de traitements adjuvants et la nécessité de stratifier les patients intérêt pour le développement d'un paysage détaillé de la physiopathologie, du diagnostic, du pronostic et de la planification du traitement de GBM. Ces thèmes relèvent essentiellement du contrôle de l'UPR et de l'IRE1 en particulier, ce que nous avons exploré dans la dernière partie de ce chapitre, où malgré la position inattaquable de l'EPU en tant que cascade de signalisation centrale dans le GBM, nous avons placé l'IRE1 sous le microscope à discerner son rôle dans GBM jusqu'à présent. Cela a permis de dégager des informations sur le fait que, même s'il existe de solides indicateurs d'implication d'IRE1 dans la GBM, il est essentiel de les caractériser davantage pour relier les concepts ci-dessus à l'activité d'IRE1. Si un tel lien était établi, IRE1 deviendrait indéniablement une cible d'intérêt pharmacologique pour la GBM.

Dans l'introduction de cette thèse, nous avons établi que l'implication de l'IRE1 dans la physiologie de la GBM devait encore être caractérisée afin d'établir sa validité en tant que cible thérapeutique ou stratifiante dans les cohortes de patients GBM. Dans le chapitre 4, nous allons procéder en disséquant de manière exhaustive la voie de signalisation en aval de IRE1, en explorant les résultats différentiels de son activité RNase, à savoir l'épissage XBP1 et le RIDD. Nous établirons des signatures de signalisation IRE1 dans le GBM et les confronterons aux données TCGA du patient, après quoi les mutations somatiques découvertes chez les patients GBM seront analysées quant à leur impact sur la signalisation IRE1. Nous explorerons l'impact de ces mutations sur le développement de la tumeur et identifierons comment la signalisation en aval de l'IRE1 peut entraîner des modifications du microenvironnement de la tumeur via les XBP1 et les RIDD. Ensuite, les résultats différentiels de ces voies seront étudiés et l'impact de RIDD sur les miARN sera étudié. À ce stade, nous verrons comment les signaux XBP1 et RIDD peuvent affecter la survie des patients et les cohortes de patients seront stratifiées en fonction de leurs signaux XBP1 ou RIDD. Après avoir présenté les résultats différentiels de l'activité IRE1 RNase dans la GBM, nous en déduirons si les lignées cellulaires primaires de GBM dérivées de tumeurs de patients réséqués peuvent récapituler les caractéristiques observées précédemment dans l'étude. Après avoir établi ces informations, nous confirmerons que de telles lignées cellulaires constituent un bon modèle de développement du développement de GBM avec une grande hétérogénéité entre les lignées cellulaires, récapitulant non seulement l’activité IRE1 différentielle, mais également les caractéristiques phénotypiques tumorales. Enfin, nous modulerons génétiquement IRE1 dans ces cellules afin de produire des lignées cellulaires stables exprimant différentes formes d’IRE1 afin d’établir un modèle cellulaire traductionnel de l’activité IRE1 dans des échantillons dérivés de patients. En tant que tel, nous tenterons de remplir le premier objectif majeur de cette thèse.

Dans ce quatrième chapitre, nous avons établi que le volet IRE1 de l'EPU joue un rôle majeur dans le développement de la GBM. En disséquant l'activité de la RNase IRE1 dans les tumeurs de GBM humain, les lignées cellulaires de GBM primaires et couramment disponibles, nous avons révélé le double rôle des XBP1 et des RIDD dans l'agressivité des tumeurs. Les tumeurs GBM se regroupent dans des groupes d'activité IRE1 élevés et faibles et les XBP1 transmettent une tumorigénicité en favorisant l'angiogenèse et le recrutement de macrophages, tandis que le RIDD atténue l'angiogenèse et la migration des cellules tumorales. Les patients porteurs de tumeurs présentant des caractéristiques XBP1s + / RIDD- ont une survie inférieure à celles des patients présentant des caractéristiques XBP1s / RIDD +, ce qui en fait de nouvelles voies thérapeutiques potentielles. Nous avons donc supervisé la réalisation du premier objectif principal en produisant un modèle GBM de traduction cellulaire de l'activité IRE1 modifiée et en établissant un rôle important pour IRE1 dans la physiopathologie de la GBM.

Dans le chapitre 4, nous avons exploré le double rôle de l'IRE1 dans la carcinogenèse des GBM et avons approfondi les détails des signaux XBP1 différentiels par rapport à la signalisation RIDD. Nous avons établi comment IRE1 agit en tant que médiateur majeur de la formation et de l'angiogenèse de tumeurs GBM, ainsi que le rôle important des miARN dans la maladie, tout en établissant un modèle cellulaire expérimental qui récapitule les tumeurs GBM tout en fournissant un état IRE1 modifié. Au chapitre 5, nous verrons plus en détail le rôle de IRE1 dans le maintien d’un phénotype de GBM spécifique. En tant que tels, nous analyserons les propriétés par lesquelles IRE1 pourrait affecter les capacités de différenciation et de reprogrammation des cellules GBM, ce qui aurait un impact profond sur les résultats cliniques car, comme il a été établi dans l’introduction, les propriétés de la tige GBM pourraient être responsables de l’évasion et de la récurrence de la GBM. Nous établirons une signature de gènes incluant des facteurs de transcription impliqués dans la reprogrammation cellulaire différenciée qui sera confrontée à une signature d'activité IRE1 établie au chapitre 4. Ensuite, les niveaux de ces facteurs de transcription seront étudiés dans des tumeurs à activité élevée et faible IRE1, soulignant ainsi les pertinence potentielle de la fonction IRE1 pour leur expression. Un modèle cellulaire de phénotype semblable à la tige et différencié sera établi dans des cellules GBM primaires, cultivées dans un milieu spécifique des lignées cellulaires adhérentes ou de la neurosphère. En utilisant les modèles cellulaires d'activité IRE1 altérée décrits au chapitre 4, nous construisons un paysage complet de reprogrammation génotypique et phénotypique en forçant la différenciation en présence ou en l'absence d'inhibition génétique ou pharmacologique de l'IRE1, en mesurant la tige, la différenciation et la reprogrammation au niveau des ARNm et des protéines. ainsi que l'observation de la capacité clonogénique et du phénotype général. En établissant un lien clair entre la signalisation IRE1 et la reprogrammation de cellules GBM, nous assisterons à une focalisation supplémentaire sur la voie de signalisation exacte en aval de IRE1 qui pourrait régir la maintenance de phénotypes différenciés et nous dévoilerons ainsi le rôle des XBP1 en tant qu'inducteur d'un sous-ensemble spécifique de miARN réguler négativement les facteurs de transcription impliqués dans la différenciation et la reprogrammation dérégulant le phénotype de maintien de la différenciation. En tant que tel, nous tenterons de remplir le deuxième objectif majeur de cette thèse. Dans ce cinquième chapitre, nous avons établi que l’axe IRE1-XBP1-miR148a est un important régulateur des facteurs de transcription en général et de SOX2 en particulier, impliqué dans la reprogrammation des cellules GBM. Nous avons combiné des approches bioinformatiques, génétiques et pharmacologiques complètes pour tester l’importance de la régulation à la baisse de IRE1 dans les systèmes cellulaires GBM. En ajoutant au répertoire de IRE1 dans la carcinogenèse des GBM, nous avons observé des découvertes qui offrent des possibilités pour la stratification des patients et le développement de thérapies combinées, tout en fournissant une raison pour la découverte de nouveaux médicaments et la modulation de IRE1 avec soin dans les GBM.

Au chapitre 5, nous avons exploré le rôle de IRE1 dans la maintenance du phénotype différencié du GBM. Au cours de cette enquête, nous avons découvert une voie de régulation des facteurs de transcription impliqués dans la reprogrammation des cellules GBM via la signalisation XBP1 et l'induction de miARN148a. Ces résultats ont renforcé la position de l’IRE1 en tant que cible pharmacologique intéressante dans la GBM, notamment dans la sensibilisation des cellules GBM aux traitements actuels. Dans le chapitre 6, nous allons essayer de faire cela et concevoir de nouveaux modulateurs IRE1 qui sensibilisent les cellules GBM à TMZ. À cette fin, nous utiliserons une nouvelle approche dans laquelle des fragments d’IRE1 seront utilisés comme matrices structurelles et de liaison pour construire des composés susceptibles de se lier in silico au domaine kinase IRE1. Étant donné que ceux-ci sont basés sur la séquence d'acides aminés IRE1, ils pourraient potentiellement porter une spécificité en raison de l'homologie de séquence et de structure de la poche de liaison de IRE1. Nous allons observer comment une bibliothèque finale de composés composés de petits peptides, de molécules approuvées par la FDA et non approuvées par la FDA a été générée puis testée in vitro et dans des systèmes cellulaires. Nous suivrons une séquence logique d'événements allant de la formation de la bibliothèque à la traduction dans un modèle GBM, en commençant par déterminer si ces composés ont affecté l'activité IRE1 in vitro dans un test de clivage expérimental propre. Après avoir établi que ces composés affectent l'activité de la RNase de la IRE1 in vitro, nous interrogerons l'effet des composés de liaison de la kinase sur la phosphorylation de la IRE1 et leur effet sur l'épissage de XBP1 en présence de deux facteurs de stress ER dans les lignées cellulaires de GBM communément disponibles et primaires. Après avoir montré qu’ils affectent effectivement l’activité IRE1 dans les cellules, nous définirons leur capacité à sensibiliser les cellules GBM au TMZ à l’aide de profils de toxicité et de doses croissantes dans les lignées primaires et couramment disponibles, combinant ainsi les modèles cellulaires décrits au chapitre 4 et le concept de chimiosensibilité proposé au chapitre 5. Ainsi, par des moyens informatiques, chimiques et biologiques, nous tenterons de remplir le troisième et dernier objectif majeur de cette thèse. Dans ce sixième chapitre, nous avons interrogé structurellement le domaine catalytique de la kinase IRE1 et avons ensuite effectué un criblage in silico qui a identifié quatre molécules approuvées et deux nouvelles molécules non approuvées cliniquement qui inhibent l'activité de la RNase de la IRE1 in vitro et en commun, ainsi que des cellules GBM primaires. . Suite à cela, la capacité des cellules GBM à faire face à TMZ en présence de ces molécules a été testée, fournissant des preuves que l'inhibition de la IRE1 kinase sensibilise les lignées cellulaires GBM humaines à TMZ. Nous avons donc supervisé la réalisation du troisième objectif principal en produisant de nouvelles molécules qui sensibilisent les cellules GBM aux thérapies actuelles. Thesis summary

This summary gives a chapter-by-chapter account of the content of the doctoral dissertation “Targeting IRE1 activity in Glioblastoma Multiform”; authored by D. Doultsinos

The first chapter of this theses serves a dual purpose. Firstly it provides the opportunity for us to take a global view of the ER as an organelle and thus examine the potential butterfly effect2 of its perturbation on normal physiology and consequently on numerous pathologies. The endoplasmic reticulum (ER) is the site of production and folding of approximately one-third of all cellular proteins; a staggering contribution to cellular and organismal life. Its smooth function is of utmost importance to life in organisms spanning archaea to humans. This functionality oscillates around a resting state which is susceptible to change and hence stress, through a multitude of factors, which could be as benign as a seasonal temperature change or the division of a cell during development. Specific signalling pathways, such as the unfolded protein response (UPR), are responsible for responding to and controlling the magnitude of these oscillations. How well they respond decides the fate of the cell making the ER and the UPR central to the balance between health and disease; life and death. As such a volume where the ER impact on other organelles and cellular behaviour and its potential exploitation to address unmet biomedical needs is of evident value. Secondly it is the examination and result of an exercise in collaborative writing as a deliverable target of a group of early stage scientists covering a variety of ER centric topics across 14 research projects carried out in 8 European research centres as part of an Innovative Training Network “TRAIN-ERS”, funded by the European Commission, through the Marie Sklodowska Curie Actions. As such we will be able to examine the benefits of interdisciplinary collaboration and subsequent results of such an undertaking through the medium of an extensive literature review. In this first chapter we have documented ER knowledge which led to its emerging roles in physiology and disease, encompassing inter-organ communication and exploitation opportunities in bioengineering and health, outcomes of which include cost effective biologics, crop or livestock improvement and the provision of precision tools to guide therapeutic innovation and stratification in specific patient pools. We have looked at how studying ER stress signalling in different organisms is enabled by its conserved nature in metazoans and therefore how, in turn, research in various organisms is complemented by high throughput and in silico strategies to produce in vitro and in vivo models, supplying us with data for pre- clinical scrutiny and subsequent clinical application. It is thus evident that understanding ER signalling and the UPR response, can enhance healthcare at a preventative, diagnostic and therapeutic level. In the next chapter we will be focusing further on the UPR and IRE1 and the ways these can be targeted.

In chapter 1 we explored the role of the ER in normal physiology and delved into the minutiae of ER signalling from a UPR centric point of view. We established how the UPR acts as a major mediator of the response of ER to physiological stress and described how UPR signalling may be exploited and is currently exploited in clinical applications. We spoke of the utilisation of in silico and high throughput techniques that allow us to build pharmacological targets and enable biomarker discovery to bring ER and UPR targeting molecules from bench to bedside, scrutinising particular examples of ER centric clinical trials and lists of UPR modulators. In chapter 2 we shall bridge the concepts of UPR signalling, molecule identification and disease application by focusing on the UPR, using specific clinical scenarios to describe its diverse cell fate decisions. As such we shall analyse its pro-death properties through the example of Amyotrophic Lateral Sclerosis (ALS) and we shall have a first look at the relationship between Glioblastoma Multiforme (GBM) and IRE1 when describing its pro-survival properties. Following this, we will summarise the ways that cellular reporters of UPR activation have been used to identify UPR modulators and in addition, we shall comb through the reporters of ER function to determine the efforts made to identify molecules targeting the secretory pathway. To complete the review of processes utilised in the uncovering of modulators, highly selective to the UPR, we will cover in silico strategies that address issues of targeted UPR modulator discovery such as compensatory mechanisms against modulation and the existence of highly homologous paralogues of the targets. Thereafter we will further focus on IRE1 activity and discuss the molecules that target it and their use in disease models. By the end of chapter 2 we shall have explored the role of the UPR, and IRE1 in particular, in disease and discussed the known processes in place that do assist in the identification and exploitation of UPR specific modulators. In this second chapter we have gone through the UPR canonical and non-canonical signalling consolidating the information acquired in chapter 1. Moreover we have revisited the main players in UPR signalling, PERK, ATF6 and IRE1 and have begun our focus on IRE1 as the most evolutionarily conserved UPR mediator shaping the way to it becoming the centrepiece of this thesis. Taking a broad perspective of the available screening strategies and reporters of ER function and UPR activation we have summarised and analysed in vitro approaches to UPR modulator discovery and synthesis. Continuing on our journey from general ER functions, to the UPR, to drug discovery to shaping the hypothesis of this thesis; IRE1 is singled out as a pharmacological target in various disease models and we have briefly explored its association with GBM pathophysiology.

In chapter 1 we explored the role of the ER in normal physiology and delved into the minutiae of ER signalling from a UPR centric point of view. We established how the UPR acts as a major mediator of the response of ER to physiological stress and described how UPR signalling may be exploited and is currently exploited in clinical applications. In chapter 2 we focused further on the role of the UPR in health and disease and began our focus into IRE1 signalling and targeting in various clinical scenarios including the main clinical focus of this thesis which is Glioblastoma multiforme (GBM). In chapter 3 we will introduce GBM as a clinical entity. Using, at first, a chronological approach, the history of GBM will be described from its first known reports to its current classification and understanding. Having gone through this initial phase that will demonstrate the leaps in knowledge over the past 200 years as well as highlight the issues associated with the complexity of characterising a disease as heterogeneous as GBM, we will delve into the clinical characteristics of the disease. Starting from a public health perspective analysing its prevalence and incidence to its established prognosis, the burden of GBM on population medicine will be analysed. Thereafter we shall account for the available clinical options to GBM patients and understand the inherent barriers to successful therapy when dealing with an invasive brain tumour, whilst providing an overview of the therapeutic landscape currently developed to be tested in GBM. Having accounted for the state of the art of therapeutic development in GBM we will further intertwine IRE1 signalling and GBM pathophysiology, providing a continuity link with chapter 2 and leading up to the rationale of why targeting IRE1 in GBM may be a valid therapeutic option based on what has already been described in chapters 1 and 2 regarding IRE1’s importance in physiology and disease and what role IRE1 has been described to have in GBM pathophysiology to date. In this third and final chapter of the introduction we have gone through the clinical concept of GBM and have established why it merits the attention that it is receiving as a clinical entity from the clinical and the translational scientific community. It is a dismal disease with much still to be discovered as to its origins and progression. After having outlined the history that has shaped the modern classification of GBM, the current therapeutic options to patients were investigated. Although it is evident that GBM research benefits from the full complement of available novel techniques and research methods it is still in dire need for further characterisation at the molecular and macroscopic level in order to achieve true patient stratification and hence improve upon treatment outcomes. Upon summarising the current promising clinical trials in GBM, focusing on phase III trials, it was evident that themes such as stemness, differentiation, drug repurposing, small molecule therapeutics, tumour sensitisation to current therapies through adjuvant treatments and need for patient stratification are of great interest in the development of a detailed landscape of GBM pathophysiology, diagnosis, prognosis and treatment planning. Such themes are very much in the sphere of control of the UPR and IRE1 in particular, something we have explored in the latter part of this chapter where despite the unassailable position of the UPR as a central signalling cascade in GBM, we put IRE1 under the microscope to discern is role in GBM so far. This has yielded information that pertains to the fact that although there are strong indicators of IRE1 involvement in GBM, further characterisation is paramount to linking the above concepts with IRE1 activity. If such link was established then IRE1 would undeniably become a target of pharmacological interest in GBM.

In the introduction of this thesis, we established that IRE1 involvement in GBM physiology needs further characterisation in order to establish its validity as a therapeutic or stratifying target in GBM patient cohorts. In chapter 4 we shall do this by comprehensively dissecting the signalling pathway downstream of IRE1, exploring the differential outcomes of its RNase activity, namely XBP1 splicing and RIDD. We will establish IRE1 signalling signatures in GBM and confront these to patient TCGA data, after which somatic mutations found in GBM patients will be analysed as to their impact on IRE1 signalling. We will explore the impact of these mutations on tumour development and pinpoint how IRE1 downstream signalling may drive changes in the tumour microenvironment through XBP1s and RIDD. Thereafter the differential outcomes of these pathways will be studied and the impact of RIDD on miRNA will be investigated. At this point we will see how XBP1s and RIDD signalling may affect patient survival and patient cohorts will be stratified according to their XBP1s or RIDD signalling. Upon presenting the differential outcomes IRE1 RNase activity in GBM we will deduce whether primary GBM cell lines derived from resected patient tumours can recapitulate the characteristics observed earlier in the study. Having established this information we will confirm that such cell lines are a good translational model of GBM development with high heterogeneity between the cell lines, recapitulating not only differential IRE1 activity but also tumoral phenotypic characteristics. Finally we will genetically modulate IRE1 in these cells to produce stable cell lines expressing different forms of IRE1 to establish a translational cellular model of IRE1 activity in patient derived samples. As such we will attempt to fulfil the first major objective of this thesis. In this fourth chapter we have established that the IRE1 arm of the UPR plays a major role in GBM development. By dissecting IRE1 RNase activity in human GBM tumours, commonly available and primary GBM cell lines we have revealed the dual role of XBP1s and RIDD in tumour aggressiveness. GBM tumours cluster in high and low IRE1 activity groups and XBP1s conveys tumorigenicity by promoting angiogenesis and macrophage recruitment whilst RIDD dampens angiogenesis and tumour cell migration. Patients bearing tumours with XBP1s+/RIDD- features have a lower survival than those with XBP1s-/RIDD+ features thereby pertaining to potential novel therapeutic avenues. We have thus overseen the completion of the first primary objective by producing a cellular translational GBM model of altered IRE1 activity and by establishing a strong role of IRE1 in GBM pathophysiology.

Mots clés : protéostase, neuro-oncologie, Keywords : proteostasis, neuro-oncology, découverte de médicament, cellules souches. drug discovery, stem cells.

Résumé : Le réticulum endoplasmique (RE) est un organite Abstract : The endoplasmic reticulum (ER) is a membranaire intracellulaire et le premier compartiment de la membranous intracellular organelle and the first voie de sécrétion. En tant que tel, le RE contribue à la compartment of the secretory pathway. As such, the ER production et au repliement d’environ un tiers des protéines contributes to the production and folding of approximately cellulaires et est donc lié au maintien de l’homéostasie one-third of cellular proteins, and is thus linked to the cellulaire. L’UPR est un processus biochimique intégré et maintenance of cellular homeostasis and the fine balance adaptatif activé en réponse au stress du RE qui contrôle between health and disease. The unfolded protein response l'homéostasie cellulaire et maintient une function (UPR) is an integrated, adaptive biochemical process that physiologique normale. L'accumulation de protéines mal controls cell homeostasis and maintains normal conformées dans le RE entraîne un stress qui peut pousser physiological function. Accumulation of improperly folded l'UPR à la signalisation apoptotique. L'UPR et l'un de ses proteins in the ER leads to stress, which may push the UPR principaux capteurs IRE1 contribuent ainsi au début, au past beneficial functions such as reduced protein production maintien et à l'exacerbation d'une multitude d'états and increased folding and clearance, to apoptotic signalling. pathologiques, y compris le glioblastome multiforme (GBM), The UPR and one of its major sensors IRE1 are thus ce qui en fait une cible thérapeutique novatrice. La GBM est contributory to the commencement, maintenance, and la tumeur primitive du système nerveux central la plus exacerbation of a multitude of disease states, including fréquente avec une incidence de 3 sur 100 000. Le pronostic Glioblastoma multiforme (GBM) making it an attractive est sombre avec des patients qui succombent à la tumeur global target to tackle conditions sorely in need of novel entre 15 et 18 mois après le diagnostic, avec une survie therapeutic intervention. médiane à 5 ans inférieure à 6%. GBM is the commonest primary CNS tumour with an Dans cette thèse, des approches in silico, in vitro et in vivo incidence of 3 per 100000. The disease has a dismal sont utilisées pour déterminer si IRE1 est un médiateur prognosis with patients succumbing to the tumour between physiopathologique majeur et une cible pharmacologique 15 and 18 months post diagnosis, with a median 5 year valide dans le GBM et si sa modulation peut fournir de survival at less than 6%. nouvelles options thérapeutiques comme traitement In this thesis, in silico, in vitro and in vivo approaches are adjuvant modifiant la maladie. utilised to assess whether IRE1 is a major Il est montré ici que IRE1 peut jouer un role différentiel dans pathophysiological mediator and valid pharmacological la physiopathologie des GBM par le biais d'angiogenèse et target in GBM and whether its modulation may provide novel de croissance, ainsi que dans l'adaptation à la therapeutic options as an adjuvant disease modifying chimiothérapie et le maintien du phénotype différencié des treatment. cellules GBM par le biais de XBP1 et de la signalisation It is here shown that IRE1 may play a differential role in RIDD. La signalisation XBP1 favorise l'infiltration des GBM pathophysiology through angiogenesis and growth as macrophages dans la tumeur, l'angiogenèse, l'invasion et le well as adaptation to chemotherapy and maintenance of maintien d'un phénotype agressif différencié, tandis que le differentiated GBM cell phenotype through XBP1s and RIDD RIDD peut atténuer les propriétés angiogénétiques et signalling. XBP1s signalling promotes macrophage invasives, ainsi que contrôler la ré-différenciation cellulaire infiltration to the tumour, angiogenesis, invasion and et l'environnement. IRE1 est évalué en tant que cible maintenance of a differentiated aggressive phenotype, whilst thérapeutique en générant des modèles cellulaires RIDD may attenuate angiogenetic and invasive properties traductionnels de GBM portant des variants génétiques as well control miRNA environment and cell re- modulés par IRE1 et en testant leur sensibilisation au differentiation. IRE1 is assessed as a therapeutic target, by témozolomide en presence d'inhibiteurs de la kinase IRE1 generating translational cellular models of GBM carrying ciblés, en établissant un nouveau pipeline de découverte de IRE1 modulated genetic variants and testing their médicaments et en produisant six médicaments non encore sensitisation to Temozolomide in the presence of targeted impactés. activité, inhibiteurs. L’ensemble de ces travaux IRE1 kinase inhibitors by establishing a novel drug discovery montre que IRE1 fait partie intégrante de la pathogenèse et pipeline and producing six as yet unknown to impact IRE1 de la progression de la GBM. Son ciblage peut s'avérer activity, inhibitors. This body of work shows that IRE1 is an bénéfique dans des sous-ensembles spécifiques de patients integral part of GBM pathogenesis and progression and atteints de GBM. targeting it may prove beneficial in specific subsets of GBM patients.