1

1

i

Université de Sherbrooke

Caractérisation fonctionnelle de l’interactome de RAB21

Par Tomas Del Olmo Programme de biologie cellulaire

Thèse présentée à la Faculté de médecine et des sciences de la santé en vue de l’obtention du grade de philosophiae doctor (Ph.D.) en Biologie cellulaire

Sherbrooke, Québec, Canada Janvier 2020

Membres du jury d’évaluation Pr Steve Jean, programme de biologie cellulaire Pr Fernand-Pierre Gendron, programme de biologie cellulaire Pr Christine Lavoie, programme de pharmacologie Pr Nicolas Bisson, Département de biologie moléculaire, de biochimie médicale et de pathologie, Faculté de médecine, Université Laval

© Tomas Del Olmo, 2020

i

ii

La recherche est une aventure qui demande beaucoup de courage.

ii

iii

1RÉSUMÉ Caractérisation fonctionnelle de l’interactome de RAB21

Par Tomas Del Olmo Programme de biologie cellulaire

Thèse présentée à la Faculté de médecine et des sciences de la santé en vue de l’obtention du diplôme de philosophiae doctor (Ph.D.) en biologie cellulaire, Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1H 5N4

Le trafic membranaire régule l’ensemble du transport vésiculaire de cargos entre les différents organites de la cellule ainsi qu’avec le milieu extracellulaire. On distingue les voies de sécrétion et d’endocytose mais celles-ci sont interconnectées. Ce trafic est un réseau complexe qui assure l’intégrité cellulaire. Une panoplie de protéines agissent donc en concert afin de permettre sa régulation. Les RAB sont considérées comme les chefs d’orchestre du trafic. Activées, elles recrutent une large gamme d’effecteurs qui permettent de médier chacune des étapes du trafic membranaire. Des défauts d’activité des RAB sont impliqués dans le développement de pathologies telles que les maladies neurodégénératives et le cancer. RAB21 est une RAB endosomale impliquée dans la régulation de l’internalisation des intégrines, du récepteur à l’EGF et dans la régulation du flux autophagique. Cependant, peu d’effecteurs et régulateurs sont connus pour RAB21. En effet, la compréhension des mécanismes moléculaires d’action des RAB reste un enjeu majeur dans le décryptage du trafic membranaire. En effet, l’identification des partenaires protéiques des RAB comporte de nombreuses limitations. Ici, la combinaison d’une approche de spectrométrie de masse quantitative (SILAC) avec une approche de marquage par proximité (APEX2) a permis d’identifier l’ensemble des partenaires protéiques de RAB21. L’approche APEX2 a permis d’identifier la plupart des interacteurs connus de RAB21 ainsi que de nouveaux effecteurs et régulateurs potentiels. La comparaison des enrichissements relatifs aux RAB5, 4, 7 et 21 a permis d’identifier des interactions spécifiques à chacune des RAB. L’approche SILAC a permis l’identification de cargos potentiels. La validation des interactions identifiées par des techniques de biochimie ont validé l’interaction de RAB21 avec les complexes WASH et Rétromère, impliqués dans des évènements de tri endosomal, ainsi qu’avec TMED10 qui régule la progression de cargos le long de la voie de sécrétion. Par la suite, la délétion de RAB21 par l’approche CRISPR/Cas9 a mis en évidence un rôle de RAB21 dans la stabilité du Rétromère et dans l’activité de WASH. La déplétion de RAB21 affecte le recyclage de cargos clathrine indépendants à la membrane plasmique. Par ailleurs, la délétion de RAB21 affecte la stabilité et la localisation au Golgi de TMED10. L’ensemble de ces travaux ont permis de caractériser de nouveaux effecteurs et cargos spécifiques ainsi que deux nouvelles fonctions impliquant RAB21. Ce projet a permis d’établir de nouvelles approches expérimentales pour élucider les mécanismes moléculaires de régulation du trafic membranaire et pourra servir de base à l’étude de l’ensemble des RAB.

Mots clés : RAB21, APEX2, SILAC, Rétromère, WASH, TMED10, trafic membranaire

iii

iv

2SUMMARY RAB21 functional interactome characterisation

By Tomas Del Olmo Cellular biology Program

Thesis presented at the Faculty of medicine and health sciences for the obtention of Doctor degree diploma philosophiae doctor (Ph.D.) in cellular biology, Faculty of medicine and health sciences, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1H 5N4

Membrane trafficking regulates vesicular transport between the different organelles of the cell as well as exchanges with the extra-cellular space. Endocytic and secretory pathways are interconnected and form a complex network assuring cell integrity. Therefore, membrane trafficking is regulated by a large set of which all act in concert to enable its proper function. RABs are considered as master regulators of trafficking. When activated, RABs recruit a wide range of effectors to mediate each steps of trafficking. Impairment of RAB activity are involved in the development of diseases such as neurodegenerative diseases and cancer. RAB21 is an endosomal RAB which regulates autophagic flux and is involved in the internalization of integrins and the EGF receptor. Despite these identified functions, only a few effectors and regulators are known for RAB21. Understanding the molecular mechanisms of action of RABs remains a major challenge to decode membrane trafficking. Indeed, identification of RABs partners with current approaches has some limitation. Here, the combination of a quantitative mass spectrometry (SILAC) approach with a proximity labeling approach (APEX2) made it possible to identify almost every RAB21 partners. The APEX2 approach allowed the identification of most of the known RAB21 interactors as well as potential effectors and regulators. The comparison of the relative protein enrichments obtained with RAB5, 4, 7 and 21 permitted to identify specific interactions. The SILAC approach has essentially allowed the identification of potential cargos for RAB21. Biochemical techniques validated the interactions of RAB21 with the WASH and Retromer complexes, both involved in endosomal sorting events, as well as with TMED10 which regulates the cargoes progression through secretory pathway. Subsequently, the deletion of RAB21 by CRISPR/Cas9 highlighted a role for RAB21 in maintaining the stability of the retromer complex and in regulating WASH complex activity. The deletion of RAB21 affected recycling of clathrin-independent-cargos to the plasma membrane. In addition, deletion of RAB21 decreased the stability of TMED10 and its location at Golgi. All this work allowed the characterization of new specific effectors and cargos as well as two new functions involving RAB21. This project is an example in the study of the molecular mechanisms of membrane trafficking regulation that can be used as a basis for the study of all RABs. Keywords: RAB21, APEX2, SILAC, Rétromère, WASH, TMED10, membrane trafficking

iv v

3TABLE DES MATIERES

1 Résumé ...... iii 2 Summary ...... iv 3 Table des matières ...... v 4 Liste des figures ...... vii 5 Liste des abréviations ...... ix 1 Introduction...... 1 1.1 Le trafic membranaire ...... 1 1.1.1 Définition ...... 1 1.1.2 La voie de sécrétion ...... 2 1.1.2.1 Le réticulum endoplasmique ...... 3 1.1.2.2 Le Golgi ...... 5 1.1.2.3 Le ERGIC ...... 6 1.1.2.4 Les vésicules COPII et COPI ...... 7 1.1.3 La famille p24 ...... 8 1.1.3.1 Les différentes protéines de la famille p24 ...... 8 1.1.3.2 Structure des protéines p24 ...... 10 1.1.3.3 Fonction des protéines p24 ...... 11 1.1.3.4 Spécificités de TMED10 ...... 11 1.2 La voie d’endocytose...... 18 1.2.1 Rôles de l’endocytose ...... 18 1.2.1.1 Endocytose clathrine dépendante ...... 19 1.2.1.2 Endocytose cavéoline dépendante ...... 21 1.2.1.3 Endocytose clathrine et cavéoline indépendante ...... 22 1.2.1.4 Endosomes précoces : gares de triage...... 24 1.2.1.5 Dégradation lysosomale ...... 26 1.2.1.6 Recyclage endosomal ...... 27 1.2.2 Le complexe Rétromère ...... 29 1.2.3 Le complexe WASH ...... 35 1.3 Les régulateurs du trafic membranaire ...... 43 1.3.1 Les RAB ...... 43 1.3.1.1 Généralités ...... 43 1.3.1.2 Structure des RAB ...... 44

v

vi

1.3.1.3 Régulation des RAB ...... 45 1.3.1.4 Les GEF ...... 47 1.3.1.5 Les GAP ...... 47 1.3.1.6 Effecteurs et fonctions des RAB ...... 48 1.3.1.7 Localisation des RAB ...... 52 1.3.2 RAB21 ...... 53 1.3.2.1 Généralités concernant RAB21 ...... 53 1.3.2.2 Localisation de RAB21 ...... 55 1.3.2.3 Fonctions de RAB21 ...... 56 1.3.2.4 Rôle de RAB21 dans des pathologies ...... 60 Hypothèse ...... 61 Objectifs ...... 62

2 Article 1 ...... 63 3 Article 2 ...... 135 4 Discussion...... 170 4.1 Comparaison des approches d’identification des partenaires de RAB21 ...... 171 4.2 RAB21 régule l’activité de WASH et du Rétromère ...... 176 4.3 Rôle de RAB21 dans la régulation de TMED10 ...... 182 4.4 Nouvelles fonctions potentielles de RAB21 ...... 185

5 Conclusion et perspectives ...... 191 6 Remerciements ...... 196 7 Liste des références ...... 197 8 Annexes ...... 217

vi vii

4LISTE DES FIGURES

Figure 1 : La voie de sécrétion du trafic endosomal ...... 4 Figure 2 : Structure des protéines de la famille p24 ...... 10 Figure 3 : Localisations et fonctions de TMED10 ...... 15 Figure 4 : Les différents types d’endocytose ...... 20 Figure 5 : L’endosome précoce, centrale de triage ...... 25 Figure 6 : Structures cristallographiques des différents manteaux formés par le Rétromère ...... 32 Figure 7 : Relation des complexes WASH et Rétromère ...... 37 Figure 8 : Régulation cyclique des RAB ...... 46 Figure 9 : Localisation des RAB ...... 53 Figure 10 : Principales fonctions associées à RAB21 ...... 57 Figure 11 : Rôle potentiel de RAB21 dans la régulation mTOR ...... 187 Article 1 Figure 1 : APEX2:RAB expression lead to endosomal biotinylation ...... 122 Article 1 Figure 2 : Comparison of APEX2:RAB21-, 5-, 4- and 7-generated proteomes .. 123 Article 1 Figure 3 : RAB21 interacts and colocalizes with the WASH and retromer complexes...... 124 Article 1 Figure 4 : Characterization of RAB21 knockout HeLa cells...... 125 Article 1 Figure 5 : RAB21 modulates WASH and retromer endosomal recruitment and WASH activity...... 126 Article 1 Figure 6 : RAB21 is required for endosomal sorting of specific cargo types. .... 127 Article 1 Figure 7 : WASH and retromer complexes are required for SLC3A2 and Basigin sorting, and for full RAB21 activation and endosomal localization...... 128 Article 1 Figure 8 : Model of RAB21-mediated WASH/retromer cargo sorting...... 129 Article 1 Figure EV 1 : Network of merged HCT116 and HeLa cells highlights new potential RAB21 interactors with roles in trafficking ...... 130 Article 1 Figure EV 2 : AP-MS identification of RAB21 interactors and Ontology enrichments of these interactors...... 131

vii

viii

Article 1 Figure EV 3 : APEX2:RAB neighbors weakly overlap with a generic endosomal probe...... 132 Article 1 Figure EV 4 : Validation of specific APEX2:RAB21 hits and further characterization of RAB21 knockout cells...... 133 Article 1 Figure EV 5 : RAB21 regulates WASH and endosomal retromer recruitment, while RAB21 knockout leads to increased SLC3A2 lysosome colocalization...... 134 Article 2 Figure 1 : RAB21 interacts with TMED10 ...... 163 Article 2 Figure 2 : RAB21 interacts with TMED10 ...... 164 Article 2 Figure 3 : TMED10 Golgi localization is altered in RAB21 knockout cells ...... 165 Article 2 Figure 4 : RAB21 knockout reduces TMED10 and TMED2 protein levels ...... 166 Article 2 Figure S 1 Indirect interaction between RAB21 and TMED10 ...... 167 Article 2 Figure S 2 TMED10 localization at trans-Golgi is modulated by RAB21 ...... 168 Article 2 Figure S 3 TMED10 stability is not affected by blocking proteasomal or lysosomal functions ...... 169 Tableau 1 : Régulateurs, effecteurs et localisations des RAB ...... 50 Tableau 2 : Partenaires protéiques de RAB21 ...... 54

viii ix

5LISTE DES ABRÉVIATIONS

ACAP1 ArfGAP With Coiled-Coil, Ankyrin Repeat And PH Domains 1 ADN Acide désoxyribonucléique ALK5 Activin Receptor-Like Kinase Receptor 5 AMBRA Activating Molecule In Beclin-1-Regulated Autophagy AMPK AMP-activated Protein Kinase ANKRD50 Ankyrin Repeat Domain 50 AP2 Adaptor Protein complex 2 APEX2 Ascorbic acid Peroxidase APP Amyloid Beta Precursor Protein ARF ADP ribosylation factor ARHGAP10 Rho GTPase Activating Protein 10 ARN Acide Ribonucléique ARP2/3 Actin Related Protein 2/3 ATG4B Autophagy Related 4B ATP7A ATPase Copper Transporting Alpha 2AR Beta-2 Adrenergic Receptor BMDM Bone Marrow Derived Macrophages CAPZ Capping Actin Protein Of Muscle Z-Line CCC COMMD/CCDC22/CCDC93 CCDC5 Coiled-Coil Domain-Containing Protein 53 CDC42 Cell Division Cycle 42 CFL1 Cofilin 1 CG CLIC/GEEC CHO Chinese Hamster Ovary CI-MPR Cation-Independent-Mannose 6-Phosphate Receptor CLIC Clathrin-Independent Carriers COPI Coat Proteins I COPII Coat Proteins II CRISPR/Cas9 Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated protein 9 CRMP2 Collapsin Response Mediator Protein-2 CSC Cargo Selective Complex CSE1L Segregation 1 Like DAF Decay-Accelerating Factor DENN Differentially Expressed in Normal and Neoplastic cells DMT1-II Divalent Metal Transporter 1 DOP1B DOP1 leucine zipper like protein B EEA1 Early Endosome Antigen 1 EGFR Epidermal Growth Factor Receptor EHD1 EH Domain Containing 1 EHD3 EH Domain Containing 3

ix

x

ERAD ER-Associated Degradation ERD2 KDEL Endoplasmic Reticulum protein retention receptor 1 ERES Endoplasmic Reticulum Exit Site ERGIC ER-to-Golgi Intermediate Compartment ESCRT Endosomal Sorting Complex Required for Transport FAK Focal Adhesion Kinase FAM21 Family with Sequence Similarity 21 FANSHY Phe-Ala-Asn-Ser-His-Tyr FCHO1/2 FCH domain only 1 FERM 4.1 protein, Ezrin, Radixin and Moesin FKBP15 FK506 Binding Protein 15 GAP GTPase Activating Protein GAPCENA GAP and Centrosome-Associated GAPVD1 GTPase Activating Protein and VPS9 Domains 1 GBF1 Golgi Brefeldin A Resistant Guanine Nucleotide Exchange Factor 1 GCN1 General Control of amino-acid synthesis 1, GDI GDP Dissociation Inhibitor GDP Guanosine Diphosphate GEEC Glycophosphatidylinositol anchored protein Enriched Endosomal Compartments GEF Guanosine Nucleotide Exchange Factor GLUT1 Glucose Transporter 1 GOLD Golgi Dynamic GPCR G Protein-Coupled Receptors GPI Glycosylphosphatidylinositol GPI-AP Glycosylphosphatidylinositol Anchored Protein GRAF1 GTPase Regulator Associated with Focal Adhesion Kinase 1 GST Glutathion S-transférase GTP Guanosine Triphosphate HOPS Homotypic fusion and Protein Sorting HRS Hepatocyte growth factor-Regulated tyrosine kinase Substrate HSC70-4 Heat Shock Protein Family A (Hsp70) Member 4 IL-1b Interleukin-1b IL-6 Interleukin-6 LC3 Light Chain 3 LPS Lipopolysaccharides LRP1 Lipoprotein Receptor-Related Protein 1 LST8 Lethal with SEC13 protein 8 MAGE-L2 MAGE Family Member L2 MCT-1 Multiple Copies in T-cell lymphoma-1 MEF Mouse Embryonic Fibroblast MICAL Microtubule Associated Monooxygenase, Calponin and LIM Domain containing 1 MICAL-L1 MICAL Like 1 MT1-MMP Membrane-Type1 Matrix Metalloproteinases MTMR13 Myotubularin-related protein 13

x xi mTOR Mammalian Target Of Rapamycin mTORC1 Mammalian Target Of Rapamycin Complex 1 mTORC2 Mammalian Target Of Rapamycin Complex 2 MVBs Multivesivular Bodies NAPA NSF Attachment Protein Alpha NFAT Nuclear Factor Activated T cell NFkB Nuclear Factor kappa-light-chain-enhancer of activated B cells NMDA N-Methyl-D-Aspartate NPF Nucleation-Promoting Factor NSF N-Ethylmaleimide Sensitive Factor N-WASP Wiskott Aldrich Syndrome protein PAR-2 Protease Activated Receptor 2 PDS5A PDS5 Cohesin Associated Factor A PDZ PSD95-Dlg1-Zo-1 PI(3)P Phosphatidylinositol-3-phosphate PI3Ks Phosphoinositide 3-kinase PI4KIIIb Phosphatidylinositol 4-Kinase Type 2 Beta PIK3R2 Phosphoinositide-3-Kinase Regulatory Subunit 2 PIKFYVE Phosphoinositide Kinase, FYVE-Type Zinc Finger Containing PIP Phosphatidylinositol Phosphate PKC Protein Kinase C Delta PLEKHM2 Pleckstrin Homology And RUN Domain Containing M2 PS1 Presenilin 1 PX Phox PYGB Glycogen Phosphorylase B RAB Ras gene from rat Brain RAC1 Ras-related C3 botulinum toxin substrate 1 RE Réticulum Endoplasmique RNF2 Ring Finger protein 2 RhoA Ras homolog gene family A RME-8 Required for Receptor-Mediated Endocytosis 8 RTK Receptor Tyrosine kinase S6K S6 Kinase SAR-1 Secretion Associated Ras Related GTPase 1 SARA Smad Anchor for Receptor Activation Sec Staphylococcal enterotoxin C SLC3A2 Solute carrier family 3 member 2 SNARE Soluble N-éthylmaleimide-sensitive-factor Attachment protein REceptor SNX Sorting Nexin SURF4 Surfeit 4 SWIP Strumpellin and WASH-Interacting Protein TBC1D Tre-Bub2/Cdc16 1 Domain Family TbRII Transforming Growth Factor Beta Receptor 2 TbVps23 Trypanosoma brucei Vaccuolar protein-associated sorting 23 TGF- Transforming Growth Factor Beta 1

xi xii

TGN Trans Golgi Network TI-VAMP Tetanus-Insensitive VAMP TLR4 Toll-Like Receptor 4 TMED Transmembrane emp24 domain-containing protein TNF Tumor necrosis factor alpha TnFR Transferin Receptor TRAPPI Transport Protein Particle I TRF2 Telomeric Repeat-binding Factor 2 TRIM27 Tripartite Motif 27 UIM Ubiquitin Interacting Motif UPR Unfolded Protein Response USP7 Ubiquitin-specific-processing protease 7 VAMP Vesicle-Associated Membrane Protein VARP Vacuolar protein sorting-associated Ankyrin-Repeat Protein V-ATPase Vacuolar-type H+-ATPase VCA Verpolin Connecting and Acidic VPS Vacuolar Protein Sorting-associated VSVG Virus de la Stomatite Vésiculeuse G WASH Wiskott Aldrich Syndrome protéine et SCAR Holomogue WASP Wiskott–Aldrich Syndrome Protein WDH WASH Domain Homology Wls Wnt ligand secretion WNT Wingless-Type XPO1 Exportin 1

xii 1

1INTRODUCTION

1.1 Le trafic membranaire

1.1.1 Définition

Chez les eucaryotes les espaces intracellulaires et extracellulaires de chaque cellule sont délimités par une bicouche lipidique : la membrane plasmique (Le Roux et al. 2019).

Cette membrane joue à la fois le rôle de barrière mais aussi d’échangeur avec le milieu extracellulaire. De plus, les cellules eucaryotes sont sous-compartimentées en unités fonctionnelles, appelées organites (Voeltz et Barr 2013). Par exemple, la réplication de l’acide désoxyribonucléique (ADN) est centralisée dans le noyau (Razin 2018), et la respiration dans la mitochondrie (Rich et Maréchal 2010). En réalité, il existe de nombreux compartiments chacun associé à une fonction. On peut noter : le cytosol, le réticulum endoplasmique lisse et rugueux, l’appareil de Golgi, les endosomes, les lysosomes, les péroxysomes, les vésicules de sécrétion, les autophagosomes, etc (Mellman et Warren 2000).

Chacun de ces compartiments est délimité par des membranes lipidiques (van Meer et al.

2008). Les cellules étant sous-compartimentées, il est donc nécessaire d’effectuer des

échanges entre ceux-ci (Costanzo et al. 2016).

L’ensemble des échanges vésiculaires entre les différents organites ainsi qu’entre l’intérieur et l’extérieur de la cellule peut être défini comme le trafic membranaire (Chavrier et Goud 1999). Le trafic s’effectue via le bourgeonnement de vésicules à partir de la membrane d’un compartiment initial, (A) (Gonzalez et Scheller 1999). Cette vésicule se détache, et migre le long du cytosquelette. Pour terminer, cette vésicule va s’ancrer et

1

2 fusionner avec la membrane d’un compartiment cible, (B) (Gonzalez et Scheller 1999).

L’ensemble des protéines présentes à la membrane et le contenu de la vésicule sont transportés du compartiment (A) vers le compartiment (B) (Pizarro et Norambuena 2014).

Le trafic est très finement régulé et des défauts de ce dernier sont impliqués dans de nombreuses pathologies, telles que les maladies neurodégénératives et le cancer (Niedergang et al. 2017). Ainsi, de nombreux complexes protéiques sont impliqués dans la régulation de chacune de ces étapes. On peut noter les protéines motrices telles les dynéines, kinésines et myosines qui jouent un rôle dans le transport des vésicules le long du cytosquelette (Hehnly et Stamnes 2007). Les protéines de manteau et SNX (Sorting Nexin) vont générer des courbures des membranes lipidiques (Gallon et Cullen 2015). Les SNARE (Soluble N-

éthylmaleimide-sensitive-factor Attachment protein REceptor), pour leur part, régulent la fusion des vésicules avec leur compartiment cible (Wang et al. 2017). Afin de maintenir une cohérence entre les différents compartiments malgré les échanges, l’enjeu est de maintenir l’identité membranaire de chaque compartiment. Cette identité est conservée par la nature des lipides, les phosphoinositides, et les protéines telles que les RAB (Ras-like in rat Brain) qui composent une membrane (Jean et Kiger 2012). On distingue deux grandes voies au sein du trafic : la voie de sécrétion et la voie de l’endocytose.

1.1.2 La voie de sécrétion

La voie de sécrétion régule les échanges de l’intérieur vers l’extérieur de la cellule

(Sallese et al. 2009). Les protéines synthétisées dans le cytosol sont transloquées au réticulum endoplasmique (RE) de façon co-traductionnelle (McCaffrey et Braakman 2016). Ces protéines sont ensuite transportées dans le compartiment ERGIC (ER-to-Golgi Intermediate

2 3 compartment) (Appenzeller-Herzog et Hauri 2006). Enfin, ces protéines transitent via l’appareil de Golgi afin d’être stockées dans des vésicules de sécrétion et être adressées à la membrane plasmique ou vers le milieu extracellulaire (Boncompain et Perez 2013).

L’ensemble des récepteurs, transporteurs, canaux, et hormones destinés à être adressés à la membrane plasmique ou sécrétés vont emprunter la voie de sécrétion.

1.1.2.1 Le réticulum endoplasmique

Le RE est impliqué dans de nombreuses fonctions dont la synthèse de lipides, de protéines et le stockage de calcium (Ca2+) (Schwarz et Blower 2016). Le RE est une structure constituée d’une bicouche lipidique prenant sa source dans la membrane externe du noyau.

Il est principalement constitué de citernes situées dans la région périnucléaire et formées par l’apposition de membranes très rapprochées (Westrate et al. 2015). En revanche, le RE s’étend à travers la cellule formant un réseau complexe grâce à l’extension de tubules pouvant créer des points de contact avec l’ensemble des autres organites de la cellule tels que la membrane plasmique, la mitochondrie, le Golgi, les endosomes et les lysosomes (Zhang et

Hu 2016). Ces points de contact jouent un rôle dans l’échange de lipides mais aussi dans la régulation d’évènements de signalisation (Dong et al. 2016; Quon et al. 2018). Le RE est le point de départ de la voie de sécrétion (Figure 1). L’ensemble des protéines sécrétées et des protéines transmembranaires sont modifiées au sein du RE. Les ribosomes synthétisant les protéines sont associés au RE, et permettent la translocation de celles-ci dans la lumière du

RE au cours de leur synthèse (Aviram et Schuldiner 2017). Lors de leur séjour dans le réticulum les protéines vont subir des changements de conformation grâce à l’intervention de protéines chaperonnes (Hartl et Hayer-Hartl 2009). Aussi, ces protéines sont soumises à des modifications post-traductionnelles telles que des N-glycosylation ou la formation de

3 4

Figure 1 : La voie de sécrétion du trafic endosomal Les protéines synthétisées dans le réticulum endoplasmique sont transférées au ERGIC via la formation de vésicules COPII aux sites ERES. Depuis l’ERGIC les cargos vont progresser vers le cis-Golgi, et finalement atteindre le réseau transgolgien. Le transport de cargos du RE vers le Golgi s’appelle le transport antérograde. Le réseau transgolgien (TGN) connecte la voie de sécrétion à l’ensemble des organites de la cellule. D’ici, les cargos sont adressés à la membrane plasmique et vers le milieu extracellulaire via les vésicules de sécrétion. Des échanges bidirectionnels sont aussi réalisés entre le TGN et les endosomes précoces, tardifs, et de recyclage. De plus, il existe un transport antérograde du Golgi au réticulum en passant par le ERGIC via la formation de vésicules COPI. Le RE représente un réseau qui s’étend à travers la cellule et réalise des contacts avec plusieurs organites ayant pour effet de réguler la scission de tubules endosomaux ou d’induire des échanges de protéines et lipides avec la membrane plasmique.

ponts disulfures (Cherepanova et al. 2016). L’ensemble de ces modifications confère leur structure tertiaire aux protéines. Ici, s’effectue un premier tri entre les protéines dont la conformation est conforme et celle dont la conformation est incorrecte. Les protéines avec des erreurs de repliement sont séquestrées dans le RE et dégradées via le système de dégradation associé au RE, ERAD (ER-associated degradation) (McCaffrey et Braakman

2016). Une fois reconfigurées et les points de contrôle validés, les protéines sont encapsidées

4 5 dans des vésicules recouvertes d’un manteau de protéines COPII (Coat proteins II), et quittent le réticulum à partir des sites de sortie du réticulum, appelés : ERES (ER-Exit Sites)

(McCaughey and Stephens 2018).

1.1.2.2 Le Golgi

Après avoir quitté le RE, les cargos transitent via le ERGIC avant d’atteindre le Golgi

(Hauri et al. 2000). Le transport du RE vers le Golgi par les vésicules COPII, s’appelle le transport antérograde. Le Golgi est constitué de multiples citernes pouvant être classées en trois sortes : le cis-Golgi qui fait face au RE et reçoit les cargos issus de celui-ci, les citernes médianes et le trans-Golgi (Wang et Huang 2017). Chacune des citernes comporte des enzymes qui permettent l’ajout de glycan sur les protéines présentes dans le Golgi (Potelle et al. 2015). Il existe deux modèles de progression des cargos à travers le Golgi. Le premier suppose que chaque citerne évolue et progresse à travers les cis-, median- et trans-Golgi

(Nakano et Luini 2010). Dans ce modèle, des vésicules COPI (Coat proteins I) contenant des protéines résidentes du Golgi, émergent depuis chaque citerne et effectuent un transport rétrograde afin de maintenir l’identité de chaque compartiment. Le second modèle propose que des vésicules COPII soient formées à chaque citerne du Golgi et permettent le transport de cargo entre les différentes parties du Golgi (Glick et Luini 2011). Une fois que les cargos ont progressé à travers le Golgi, ils atteignent le réseau transgolgien (TGN). C’est au TGN que sont formées les vésicules de sécrétions qui sont adressées à la membrane plasmique permettant ainsi la sécrétion de molécules vers le milieu extracellulaire mais aussi l’adressage de protéines transmembranaires et de lipides à la membrane plasmique (Guo, Sirkis, and

Schekman 2014). Le TGN constitue la gare de triage de la voie de sécrétion (Figure 1). Des

échanges entre le Golgi et les différents compartiments sont réalisés (Boncompain et Perez

5 6

2013). En effet, il existe des échanges membranaires bidirectionnels entre le Golgi et les endosomes précoces, de recyclage et tardifs. Il existe aussi un transport rétrograde via des vésicules constituées de protéines de manteau COPI, du Golgi vers le RE (Lee et al. 2004).

1.1.2.3 Le ERGIC

Les échanges entre le RE et le Golgi sont très dynamiques et bidirectionnels. Entre ces deux compartiments existe un compartiment intermédiaire, le ERGIC (Appenzeller-Herzog et

Hauri 2006). Formé de l’accumulation de vésicules COPII et COPI transportant des cargos entre le RE et le Golgi, l’ERGIC constitue un compartiment facilitant les échanges entre le

RE et le Golgi mais aussi un compartiment supplémentaire de contrôle de la progression des cargos le long de la voie de sécrétion (Figure 1). En effet, il existe deux hypothèses concernant le fonctionnement du ERGIC (Appenzeller-Herzog et Hauri 2006). La première considère le transport antérograde de l’ERGIC le long des microtubules. L’ERGIC

émergerait de la fusion de vésicules COPII, migrerait le long des microtubules et fusionnerait avec le cis-Golgi, participant ainsi à sa formation (Stephens et Pepperkok 2001). La formation de vésicules COPI émergeant du Golgi, et leur transport rétrograde le long des microtubules, dans ce cas, permettent le recyclage de composants résidents du RE. La deuxième hypothèse considère que l’ERGIC est un compartiment stable (Presley et al. 1997).

Dans ce cas, les vésicules COPII émergeant du RE migreraient vers l’ERGIC indépendamment des microtubules puis de l’ERGIC vers le Golgi via les microtubules. Ici, le recyclage du Golgi vers l’ERGIC serait aussi dépendant de vésicules COPI migrant le long des microtubules.

6 7

1.1.2.4 Les vésicules COPII et COPI

Les vésicules COPII et COPI régulent l’ensemble du trafic, à savoir la progression et le recyclage des cargos entre le RE et le Golgi (Béthune et Wieland 2018). Ces vésicules sont dénommées ainsi en fonction du manteau protéique qui les constitue. Les vésicules

COPII naissent au sites ERES lors de l’accumulation de cargos (Langhans et al. 2012). Des protéines de contrôle qualité sont présentes à ces sites et permettent d’exclure les protéines avec des défauts de conformation. Le bourgeonnement de ces vésicules est initié par SAR-1

(Secretion Associated Ras Related GTPase 1) qui est activée par Sec12 (Weissman, et al.

2001). Sec12 étant une protéine résidente du RE, cela permet de limiter l’activité de SAR-1 au RE (Kurokawa, et al. 2016). Une fois activée, SAR-1 est ancrée dans la bicouche lipidique du RE. Ensuite, SAR-1 recrute l’hétérodimère Sec23/Sec24 qui constitue le pré-manteau des vésicules COPII (Matsuoka et al. 1998). Sec23 interagit avec SAR-1 alors que Sec24 interagit directement avec les cargos présents à la membrane des ERES (Mancias and Goldberg 2008).

Il existe plusieurs isoformes de Sec24 ce qui permet la sélection et le transport d’une grande variété de cargos (Wendeler et al. 2007). Enfin, l’hétérodimère Sec13/Sec31 est recruté via l’interaction entre Sec23/Sar1 et Sec31 (Bi et al. 2007). Sec13/Sec31 forment une cage qui induit une courbure de la membrane et participe à la fission de la vésicule (Stagg et al. 2006).

Ainsi, la vésicule est transportée le long des microtubules puis dénudée de son manteau

COPII via l’inactivation de SAR-1. Finalement, la fusion de cette vésicule avec la membrane du ERGIC permet la progression des cargos le long de la voie de sécrétion (Peotter et al.

2019). Les vésicules COPI diffèrent par la composition de leur manteau, et réalisent le trajet inverse. En effet, elles sont formées au Golgi ou à l’ERGIC. L’activation de ARF1 (ADP ribosylation factor 1) par GBF1 (Golgi Brefeldin A Resistant Guanine Nucleotide Exchange

7 8

Factor 1) permet le recrutement du complexe COPI formé d’un hétéro-heptamère (Monetta et al. 2007). Ces sept protéines peuvent être sous-divisées en deux sous-complexes. Le sous- complexe F, composé des sous-unités , ’ et -COP, et le sous-complexe B, formé des sous- unités , ,  et -COP (Bykov et al. 2017). De la même manière que COPII, le manteau

COPI va permettre le bourgeonnement et la fission d’une vésicule permettant son transport le long des microtubules vers le RE (Arakel et Schwappach 2018). L’inactivation d’ARF1 va faciliter le détachement du manteau COPI et permettre la fusion de la vésicule avec la membrane du RE, réalisant le transport rétrograde de cargo (Presley et al. 2002). Les sous- unités -COP, ’-COP ainsi que -COP et -COP ont la capacité d’interagir avec différents motifs des cargos (Bremser et al. 1999). La sélection des cargos aux vésicules COPI ou

COPII représente une étape essentielle dans le trafic des cargos afin qu’ils soient triés vers le bon compartiment. Il a été montré que les protéines de la famille p24 jouent un rôle essentiel dans la formation des vésicules COPI par le recrutement d’ARF1 mais aussi dans la sélection des cargos puisque les protéines p24 interagissent directement avec le motif KDEL qui est un signal de transport rétrograde vers le RE (Irina Majoul et al. 2001).

1.1.3 La famille p24

1.1.3.1 Les différentes protéines de la famille p24

Avec 11 membres chez les vertébrés, ces protéines sont impliquées dans la régulation du trafic des cargos entre le RE et le Golgi (Strating et Martens 2009). Les membres de la famille p24 sont toutes des protéines transmembranaires de type I d’environ 24 kD qui ont

été classées selon leur homologie en acide aminé. Elles sont séparées en quatre sous catégories , ,  et  (Bremser et al ). Les protéines p24 sont exprimées chez

8 9 l’ensemble des eucaryotes (Pastor-Cantizano et al. 2016). En revanche, la conservation des sous catégories varie selon les espèces. L’expression de la plupart des protéines de la famille p24 est ubiquitaire chez les vertébrés malgré des niveaux variables selon le tissu, ceci pouvant

être expliqué par la nécessité de médier le transport de cargos spécifiques (Rötter et al. 2002).

En effet, les protéines p24 sont hautement exprimées dans les cellules sécrétrices comme les cellules endocrines. Par exemple, la forte expression des p24 dans les cellules pancréatiques sont associées à une forte sécrétion d’insuline (L. Zhang et Volchuk 2010). Il existe les deux exceptions p24α1 and p24γ5 qui chez la souris semblent avoir un profil d’expression restreint au pancréas et aux poumons, foie, rein et tractus intestinal, respectivement (Strating et al.

2009). Les protéines p24 sont localisées sur l’ensemble des organites associés à la voie de sécrétion c’est-à-dire au RE, à l’ERGIC et au Golgi (Füllekrug et al. 1999). Des sous populations ont aussi été identifiées à la membrane des péroxysomes, aux vésicules de sécrétion et à la membrane plasmique (Marelli et al. 2004; Langhans et al. 2008). Par ailleurs, les protéines p24 sont associées aux vésicules COPI et COPII (Montesinos et al. 2014; Muñiz et al. 2000). L’ensemble des membres possèdent un motif dibasique permettant de lier les sous unités Sec23/Sec24 du manteau COPII (Belden et Barlowe 2001). De même toutes les protéines de la famille p24 ont la capacité d’interagir avec la sous-unité γ-COP des vésicules

COPI (Béthune et al. 2006). Bien que les mécanismes soient méconnus, les sous-catégories

 et  arborent un motif dilysine ayant été caractérisé comme étant important (L. P. Jackson et al. 2012). Ainsi les protéines p24 peuvent cycler entre le RE et le Golgi via le transport médié par les vésicules COPI et COPII.

9 10

1.1.3.2 Structure des protéines p24

L’ensemble des protéines p24 possède la même structure. Elles sont constituées d’un court domaine cytosolique de 13 à 20 acides aminés en C-terminal (Pastor-Cantizano et al.

2016). Les motifs dilysine et dibasique sont présents sur ce domaine (Figure 2). Un domaine transmembranaire permet l’intégration dans la bicouche lipidique. Dans la lumière des organites, les p24 présentent un domaine en superhélice, suivit d’un domaine d’association et d’un domaine GOLD (Golgi dynamique) (Anantharaman et Aravind 2002). Le domaine en superhélice est impliqué dans l’oligomérisation de protéines p24 (Ciufo et Boyd 2000).

En effet, des expériences de co-immunoprécipitations et de purifications GST (Glutathion S- transférase), ont montré qu’il existe différentes combinaisons d’association entre les membres de la famille p24, formant des hétérodimères et hétérotétramères (Fujita et al.

2011). Par ailleurs, il a été montré que l’inhibition de l’un des membres de la famille p24 entrainait la réduction de l’expression des autres membres, soutenant le fait qu’ils soient associés en complexes (D’Arcangelo et al. 2015).

Figure 2 : Structure des protéines de la famille p24

Les protéines de la famille p24 possèdent quatre domines distincts. Un court domaine cytosolique en C-terminal possède le motif ΦFXXBBX de liaison aux protéines des manteaux COPI et COPII. Φ est un résidu hydrophobique, et BB sont deux groupes basiques. Le domaine transmembranaire (TMD) permet l’insertion dans la bicouche lipidique et l’association avec d’autres protéines de la famille p24. Les protéines p24 possèdent deux domaines dans la lumière du l’organite. Le domaine CC (coil-coiled) est séparé du domaine GOLD (Golgi dynamique) et TMD par deux courtes séquences liens. Figure tirée de Pastor- Cantizano et al. 2016. Autorisation d’emprunt de la figure obtenue de Springer Nature.

10 11

Enfin, le domaine GOLD joue un rôle dans les interactions protéine-protéine et dans la reconnaissance de cargos (Ciufo et Boyd 2000).

1.1.3.3 Fonction des protéines p24

Initialement identifiées comme jouant un rôle dans le trafic de cargos associés à une ancre GPI (glycosylphosphatidylinositol), il a été montré que l’expression des membres de la famille p24 est impliquée dans le maintien de l’intégrité du Golgi (Emery et al. 2000). De même, l’inhibition de leur expression induit la réponse UPR (unfolded protein response), réponse du stress au RE (Belden et Barlowe 2001). Les protéines p24 sont aussi impliquées dans la formation des vésicules COPI (Aguilera-Romero et al. 2008). En effet, ARF1 est recruté au Golgi par interaction directe avec les domaines cytosoliques du dimère

TMED10/TMED2 (Transmembrane emp24 domain-containing). Ensuite, les p24 interagissent avec les sous-complexes F et B, favorisant la formation des vésicules COPI

(Pastor-Cantizano et al. 2016; Popoffe et al. 2011). De plus, les protéines p24 possèdent un domaine cytosolique permettant le recrutement et l’intégration dans les vésicules COPI d’ERD2 (KDEL Endoplasmic Reticulum protein retention receptor 1), un récepteur des cargos possédant un motif KDEL (Montesinos et al. 2014). Enfin les protéines p24 permettent l’inhibition d’ARF1, étape nécessaire à démanteler le manteau COPI (Majoul et al. 2001).

1.1.3.4 Spécificités de TMED10

TMED10 (Transmembrane Emp24 Domain), aussi appelé p231 ou encore p23 et

Tmp21, est l’un des membres de la famille p24 qui a été le plus étudié. Chez les humains, l’expression de TMED10 est induite par le facteur de transcription NFAT (Nuclear Factor

11 12

Activated T cell) (S. Liu et al. 2011). TMED10, l’un des deux membres constituant la sous- catégorie  chez les vertébrés est conservée des mammifères à la levure (Pastor-Cantizano et al. 2016). TMED10 est présent dans la plupart des tissus avec des niveaux qui varient puisque

TMED10 est fortement exprimé dans les reins, de manière modérée dans le cerveau, et faiblement dans les muscles (Strating et al. 2009). De plus, l’expression de TMED10 est modulée au cours du développement. En effet, TMED10 est fortement exprimé dans le cerveau dans les étapes précoces du développement et voit son expression diminuer au cours de la vie de l’individu (Vetrivel et al. 2008). Par ailleurs, la déplétion de TMED10 chez la souris est homozygote léthale au stade embryonnaire (Denzel et al. 2000). Ainsi, TMED10 semble jouer un rôle au cours du développement. Enfin, TMED10 est régulé par des modifications post-traductionnelles. En effet, selon les modèles cellulaires TMED10 peut

être glycosylé (Osiecka-Iwan et al. 2014). TMED10 dans des cellules humaines possède une demi-vie de trois heures. Les niveaux protéiques de TMED10 sont régulés par sa dégradation au protéasome en réponse à son ubiquitination (S. Liu et al. 2008).

De la même manière que les autres membres de la famille p24, TMED10 forme des hétérodimères. Des expériences de co-immunoprécipitation montrent que TMED10 interagit avec TMED2 dans des cellules CHO (Chinese hamster ovary) (Gommel et al. 1999). Dans les cellules HeLa, il a été montré par co-immunoprécipitation que TMED10 pouvait interagir avec TMED7 (Füllekrug et al. 1999). Enfin, des expériences de purification GST dans des cellules CHO montrent des interactions entre TMED10 et les membres TMED9, TMED5 et

TMED2 (Fujita et al. 2011). Pour soutenir la formation d’hétérodimères fonctionnels, il a été montré que l’inhibition de l’expression de TMED10 entraine une réduction des niveaux d’expression de TMED9 et TMED2 et ce, dans des lignées cellulaires pancréatiques (Denzel

12 13 et al. 2000). Aussi, la déplétion de TMED10 dans des cellules CHO induit la réduction des niveaux d’expression de TMED9, TMED2 et TMED5 (Zhang et Volchuk 2010).

Inversement, des souris mutantes déplétées pour TMED2 montrent des niveaux d’expression de TMED10 réduits (Theiler et al. 2014). La déplétion de TMED10 entraine une réduction des niveaux d’expression des isoformes TMED1, TMED7, et TMED3 (Vetrivel et al. 2007).

L’ensemble de ces données suggèrent que TMED10 peut former plusieurs hétérodimères et/ou des tétramères.

Comme les autres protéines de la famille p24, TMED10 est essentiellement localisé au cis-Golgi, au RE et à l’ERGIC (Blum et al. 1999). Des données montrent que la localisation de TMED10 dépend de son domaine superhélice (coil-coiled) intraluminal et des dimères formés avec les différents membres de la famille p24 (Emery et al 2000). Cependant, d’autres résultats montrent que les domaines transmembranaire et cytosolique sont suffisants pour le trafic de TMED10 entre le RE et le Golgi, alors que le domaine intraluminal ne semble pas nécessaire (Blum et Lepier 2008). En revanche, le domaine luminal est nécessaire au transport du Golgi à la membrane plasmique.

Il a été montré que TMED10 forme un complexe avec TMED2 transitant entre le ER et le Golgi. Le complexe TMED10/TMED2 est particulièrement enrichi au niveau de vésicules COPI (Gommel et al. 1999). En effet, le domaine cytosolique de TMED10 interagit avec ARF1 et les sous-complexes B et F du manteau COPI permettant leur recrutement à la membrane du Golgi et régulant ainsi la formation des vésicules COPI (Bremser et al. 1999).

De même, TMED10 interagit avec le manteau COPII mais avec une affinité moindre

(Dominguez et al. 1998). L’impact de TMED10 ne se limite pas à la formation des vésicules

COPI, puisque la perte et la surexpression de TMED10 affectent la structure du Golgi. Par exemple, la surexpression de TMED10, dans des cellules neuronales N2, entraine une

13 14 fragmentation du Golgi (Gong et al. 2011). Enfin, TMED10 est impliqué dans le transport de cargos associés à une ancre GPI (Figure 3). L’inhibition de TMED10 dans les cellules

CHO induit un retard de transport à la membrane plasmique de la protéine virale VSVG

(Vesicular Stomatitis Virus G) seulement lorsque celle-ci est liée à une ancre GPI (Takida et al. 2008). De même, la déplétion de TMED10 entraine des retards de maturation d’autres protéines possédant une ancre GPI telles que DAF (Decay-Accelerating Factor), et CD59

(Takida et al. 2008). Aussi, il a été montré que TMED10 par l’intermédiaire de TMED2 permet la séquestration du récepteur PAR-2 (Protease Activated Receptor 2) au niveau du

Golgi lorsque les cellules ne sont pas stimulées (Luo et al. 2011).

Les fonctions de TMED10 ne se limitent pas à la régulation du trafic membranaire entre le RE et le Golgi (Figure 3). Certes, TMED10 est localisé au cis-Golgi mais des sous- populations sont aussi présentes à la membrane des vésicules de sécrétion dans les cellules  pancréatiques (Zhang et Volchuk 2010). Ainsi, l’inhibition de TMED10 inhibe la sécrétion d’insuline en réponse à des stimulations en glucose.

Aussi, des évidences montrent que TMED10 est impliqué dans des processus tumorigènes. Suivant les mêmes types de mécanismes moléculaires impliqués dans la rétention de PAR-2 au Golgi, l’interaction de TMED10 avec le domaine C1 de la PKC

(Protein Kinase C) favorise la séquestration de celle-ci à des compartiments périnucléaires, et ce, prévenant sa localisation à la membrane plasmique (Wang et al. 2011). La réduction de l’expression de TMED10 par une approche d’ARN (Acide Ribonucléique) interférence

(ARNi) a donc pour effet de favoriser l’accumulation de la PKC à la membrane plasmique.

De plus, la déplétion de TMED10 est associée à une activation de la PKC et une induction de l’apoptose. Par ailleurs, une étude montre que l’expression de TMED10 est impliquée

14 15

Figure 3 : Localisations et fonctions de TMED10 TMED10 est principalement localisée entre le réticulum endoplasmique et le Golgi. TMED10 régule la formation de vésicules COPI et COPII. TMED10 est aussi impliquée dans la progression de cargos avec une ancre GPI et est donc présente sur tous les compartiments le long de la voie de sécrétion du réticulum endoplasmique à la membrane plasmique. TMED10 est aussi localisée aux vésicules d’endocytose. Enfin, TMED10 permet la rétention de cargo au réticulum endoplasmique tels que PAR-2 et la PKC

15 16 dans l’inhibition de la voie de signalisation mTOR (Xu et al. 2015). La voie mTOR est impliquée dans l’inhibition de l’autophagie et l’induction de la synthèse protéique jouant un rôle dans la croissance cellulaire (Wei et al. 2019). La voie mTOR est suractivée dans la plupart des types de cancer. La réduction de l’expression de TMED10 a pour effet de réduire les niveaux de phosphorylation de mTOR sur la Ser2448 et de S6K sur la Thr389 qui est une cible de mTOR. De plus, la déplétion de TMED10 active la phosphorylation de la AMPK

(AMP-activated Protein Kinase) et augmente le ratio LC3-II/LC3-I (Light Chain 3), synonyme d’une induction de l’autophagie (Xu et al. 2015). Finalement, la croissance cellulaire est augmentée lors de la surexpression de TMED10 et inhibée lors de sa déplétion.

Ces deux études semblent montrer que TMED10 est impliquée dans l’inhibition de la mort cellulaire et la promotion de la survie et de la croissance. En revanche, TMED10 a aussi pour effet d’inhiber la signalisation TGF- (Transforming Growth Factor ). En effet, le domaine extracellulaire de TMED10 interagit avec les deux récepteurs au TGF de type I

(ALK5) et de type II (TRII) (Nakano et al. 2017). L’inhibition de l’association entre ALK5

(Activin Receptor-like Kinase receptor 5) et TRII (Transforming Growth Factor Beta

Receptor 2) induite par la surexpression de TMED10 a pour effet d’inhiber la signalisation

TGF-. En effet, l’expression du peptide correspondant au domaine cytosolique de TMED10 suffit à inhiber la phosphorylation et l’activité du facteur de transcription Smad2 qui est impliqué dans la signalisation TGF- Finalement, dans des modèles de xénogreffes de cellules cancéreuses JygMC dont la croissance dépend de la voie TGF-, la surexpression de

TMED10 a pour effet de réduire la taille de la tumeur (Nakano et al. 2017). Dans ce même contexte la déplétion de TMED10 induit la croissance des tumeurs. Selon le contexte

16 17 cellulaire TMED10 semble avoir des effets ambivalents concernant la régulation de la survie et de la croissance cellulaire.

De plus, TMED10 semble jouer un rôle dans le développement de la maladie d’Alzheimer

(Shin et al. 2019). Le premier indice vient de l’observation que l’expression de TMED10 est fortement réduite dans le cerveau des personnes atteintes de la maladie d’Alzheimer (Shin et al. 2019) alors que l’expression de TMED10 diminue dans le cerveau au cours du vieillissement (Vetrivel et al. 2008). L’un des symptômes de la maladie d’Alzheimer est l’accumulation de plaques d’amyloïdes- L’amyloïde- est formé par le clivage successif de la protéine APP (Amyloid Beta Precursor Protein) par la - et la -sécrétase (Selkoe 2001).

Ainsi des études montrent que l’inhibition de TMED10 favorise la stabilité et la sécrétion de l’APP ainsi que l’augmentation de la phosphorylation de l’APP favorisant le clivage par la

-sécrétase. Ceci ayant pour résultat une accumulation des amyloïde  et  dans des neuroblastomes (Pardossi-Piquard et al. 2009). Enfin, La déplétion de TMED10 induit l’autophagie. TMED10 interagit avec ATG4B (Autophagy Related 4A) qui est impliqué dans le clivage de LC3. Lors de carence, l’interaction entre TMED10 et ATG4B est inhibée, ce qui a pour effet d’induire le clivage de LC3 et d’activer l’autophagie. Or, l’autophagie a un rôle protecteur dans la maladie d’Alzheimer car elle permet la dégradation de l’amyloïde-

 (Li et al. 2017) Ici aussi, le rôle de TMED10 dans le développement de cette pathologie semble complexe. D’une part, son expression semble prévenir l’accumulation des amyloïde

 et  et d’autre part TMED10 semble inhiber l’autophagie.

Pour résumer, TMED10 est un membre de la famille p24 qui joue un rôle essentiel dans la voie de sécrétion et semble être impliquée dans chacune des étapes, du trafic de protéines du RE à la membrane plasmique (Figure 3). Enfin, TMED10 joue un rôle particulier

17 18 dans la formation de vésicules COPI et COPII. Par son domaine intraluminal TMED10 est impliquée dans la régulation de plusieurs voies de signalisation en jouant sur la séquestration et la sécrétion de différents cargos. De plus, son domaine cytosolique a aussi la capacité de réguler des voies de signalisation par interaction directe avec des récepteurs membranaires.

Ainsi, TMED10 joue des rôles importants dans la régulation des processus tumorigènes en modulant l’activité des voies mTOR, TGF-, PKC et de l’autophagie. Enfin, TMED10 est importante dans le développement de la maladie d’Alzheimer. En revanche, les mécanismes moléculaires impliqués dans le recrutement de TMED10 en amont de la formation des vésicules COPI et COPII, et en particulier ceux impliqués dans le triage de la sous population présente à la membrane plasmique sont méconnus.

1.2 La voie d’endocytose

1.2.1 Rôles de l’endocytose

Le trafic endosomal permet l’internalisation et le transport vésiculaire de macromolécules issues du milieu extracellulaire et de protéines présentes à la membrane plasmique vers l’ensemble des compartiments cellulaires (Elkin et al. 2016). En effet, l’invagination de la membrane plasmique et la formation de vésicules permet de capturer et de transporter des cargos au sein de sphères lipidiques appelées vésicules d’endocytose (Roth et Porter 1964). Les vésicules nouvellement formées sont alors transportées et incorporées par fusion des membranes aux endosomes précoces (Helenius et al. 1983). Les endosomes précoces peuvent aussi fusionner entre eux, maturer et finalement fusionner avec les lysosomes (Huotari et Helenius 2011; Luzio et al. 2000). Une fois aux endosomes précoces, les cargos, en fonction des besoins cellulaires, sont triés vers les différents organites

(Parachoniak et Park 2012). En effet, ils peuvent être recyclés à la membrane plasmique,

18 19 transportés au réseau transgolgien ou encore triés vers la dégradation lysosomale (Elkin et al. 2016). L’intégrité du trafic endosomal joue un rôle essentiel dans l’homéostasie membranaire et dans la régulation de nombreuses voies de signalisation (Blazek et al. 2015;

Lobingier et von Zastrow 2019). Ce trafic est extrêmement complexe et hautement régulé.

De nombreux complexes protéiques ainsi que différentes classes de lipides sont coordonnés afin d’assurer le maintien du trafic endosomal (Lodish et al. 2000; Blazek et al. 2015). Il existe plusieurs types d’endocytose avec chacune ses spécificités (Mayor et al. 2014;

Kaksonen et Roux 2018). Le type d’endocytose influence aussi le destin des cargos endocytés.

1.2.1.1 Endocytose clathrine dépendante

Le type d’endocytose la plus étudiée est l’endocytose clathrine dépendante (Figure 4), appelée ainsi en raison du manteau de clathrine présent à la surface de ces vésicules

(McMahon and Boucrot 2011). Les vésicules de clathrine sont formées soit de manière aléatoire à la membrane plasmique soit de manière induite. En effet, l’activation de récepteurs couplés aux protéines G (RCPG) ou des récepteurs à activité tyrosine kinase (RTK) peut induire l’endocytose (Goh et Sorkin 2013; Pavlos et Friedman 2017). L’endocytose peut être découpée en plusieurs étapes successives : l’initiation, la sélection des cargos, la croissance, la fission et enfin le relargage du manteau (Gallon et Cullen 2015; Kaksonen et Roux 2018;

Jackson 2014). L’initiation se fait par le recrutement de protéines adaptatrices telles que AP2

(adaptor protein 2), FCHO1/2 (FCH domain only 1/2) qui interagissent avec des récepteurs

à la membrane plasmique ainsi qu’avec le PI(4,5)P2 (Henne et al. 2010).

19 20

Figure 4 : Les différents types d’endocytose Les différents types d’endocytose peuvent être classés selon la taille des vésicules, le type de manteau et les mécanismes moléculaires impliqués. La phagocytose permet à la cellule d’internaliser des agents pathogènes et de les dégrader par la voie lysosomale. La macropinocytose permet la capture de composants extracellulaires par l’extension de la membrane plasmique. Ces deux types d’endocytose forment des vésicules de l’ordre de 1m. Il existe les endocytoses impliquant la dynamine. Parmi ces endocytoses, on distingue les vésicules qui arborent soit un manteau de clathrine (120 m) soit de cavéolines (60 m) ou qui dépendent de RhoA (90 m). Enfin, il existe trois types d’endocytose qui ne dépendent pas de la dynamine mais de GRAF1, ARF6 ou CDC42, formant des vésicules de l’ordre de 90 m. Tout type d’endocytose dépend de la polymérisation d’actine par les complexes N- WASP et ARP2/3. Enfin, toutes les vésicules formées vont fusionner avec les endosomes précoces. Les structures CLIC/GEEC (Clathrin-Independent Carriers/ GPI-AP-enriched early endosomal compartments) résultent de la fusion des vésicules d’endocytoses indépendantes de la clathrine et de la dynamine.

Les protéines adaptatrices recrutent ensuite la clathrine. De la même manière que pour les vésicules COPI et COPII, l’assemblage de clathrine induit une courbure de la membrane plasmique et le bourgeonnement de la vésicule (Ungewickell et Hinrichsen 2007). La sélection des cargos se fait via l’interaction entre le domaine cytosolique de protéines transmembranaires et les protéines adaptatrices présentes dans le manteau telles que les

Epsines et le complexe AP-2 (Kazazic et al. 2009; Jackson et al. 2003). Les mécanismes de

20 21 croissance et de fission sont régulés par le recrutement du complexe WASP (Wiskott–Aldrich

Syndrome protein) qui médie la polymérisation d’active via ARP2/3 (Actin Related Protein

2/3) (Alekhina, et al. 2017). La polymérisation d’actine va induire une tension qui favorise le détachement de la vésicule d’endocytose. La fission est modulée par l’action de la dynamine qui étrangle la membrane de la vésicule à sa base en formant un tubule étroit et permettant le détachement de la vésicule (Antonny et al. 2016). Enfin, le manteau de clathrine se détache par l’effet de HSC70-4 (Heat Shock Protein Family A (Hsp70) Member 4) qui interagit avec le manteau de clathrine et déstabilise l’intégrité du maillage (Xing et al. 2010).

1.2.1.2 Endocytose cavéoline dépendante

La clathrine n’est pas le seul type de manteau impliqué dans la voie d’endocytose

(Figure 4). Il existe plusieurs endocytoses indépendantes des clathrines, dont l’endocytose dépendante des cavéolines (Pelkmans et Helenius 2002). Les cavéolines sont des protéines ancrées et adressées à la membrane plasmique via la voie de sécrétion décrite précédemment.

Les cavéolines arborent une structure avec un domaine cytosolique tige-boucle (Stan 2005).

Les cavéolines recrutent les cavines, protéines qui participent à la formation du manteau et l’induction des courbures de la membrane (Hill et al. 2008). Les cavéolines ont la spécificité d’être enrichies aux niveaux de radeaux lipidiques, et peuvent interagir avec le cholestérol

(Lajoie et Nabi 2010). Il a été montré que les cavéolines sont impliquées dans l’internalisation des glycosphingolipides et d’acides gras au sein de la cellule (Örtegren et al. 2004). Par ailleurs, le domaine cytosolique des cavéolines peut jouer le rôle de plateforme de signalisation indépendamment de leur rôle dans l’endocytose (Martinez-Outschoorn et al.

2015). De la même manière que pour l’endocytose dépendante de la clathrine, l’endocytose

21 22 via les cavéolines nécessite l’intervention de la dynamine, et de la polymérisation d’actine afin d’induire le détachement de la vésicule (Henley et al. 1998).

1.2.1.3 Endocytose clathrine et cavéoline indépendante

L’endocytose clathrine et cavéoline indépendantes peut être classifiée en sous- catégories selon la taille des vésicules mais aussi l’intervention de la dynamine (Mayor et al.

2014). Les endocytoses permettant l’internalisation de larges structures membranaires, supérieures à 200 nm, sont la macropinocytose et la phagocytose (Lim et Gleeson 2011),

(Flannagan et al. 2012). Ces types d’endocytose utilisent des machineries communes aux autres classes, et en particulier le réarrangement du cytosquelette d’actine pour remodeler la membrane plasmique (Figure 4). Les endocytoses dépendantes de la dynamine sont caractérisées par le type de cargos endocytés. Par exemple, l’accumulation du récepteur à l’interleukine II dans des radeaux lipidiques permet l’activation de RhoA (Ras homolog gene family A) puis Rac1 (Ras-related C3 botulinum toxin substrate 1) qui induit le recrutement du complexe N-WASP et ARP2/3 (Lamaze et al. 2001). La polymérisation d’actine facilite alors l’internalisation du récepteur de l’interleukine. L’internalisation de ces cargos se produit par leur accumulation dans des domaines membranaires riches en cholestérol, ce qui induit l’internalisation de ceux-ci via la formation de vésicules sous l’effet du complexe N-

WASP et de RhoA (Liu et al. 2009). Il existe ensuite les voies d’endocytose ne dépendant ni de la clathrine ni de la dynamine (Figure 4). Ces voies sont associées à des structures riches en protéines possédant une ancre GPI (Elkin et al. 2016). Des vésicules ou des tubules d’endocytose émergent de la membrane plasmique sans protéine de manteau, structures appelées CLIC (Clathrin-Independent Carriers), fusionnent ensemble et forment des structures riches en GPI-AP, nommées GEEC (GPI-Enriched Endocytic Compartments). Ces

22 23 endocytoses clathrine indépendantes, CLIC et GEEC, sont nommées CG. Cette endocytose est induite par l’activation à la membrane plasmique d’ARF1 par GBF1 (Monetta et al. 2007).

ARF1 activée, elle recrute ARHGAP10 (Rho GTPase Activating Protein 10), un inhibiteur de CDC42 (Cell Division Cycle 42) (Kumari et Mayor 2008). Cela permet de maintenir

CDC42 dans son cycle d’activation, et donc de permettre la polymérisation d’actine à la membrane plasmique. Une autre protéine importante pour cette endocytose est GRAF1

(GTPase Regulator Associated with Focal Adhesion Kinase 1) qui détient un domaine BAR pouvant induire des courbures de la membrane plasmique (Hansen et Nichols 2009). Les cargos endocytés via la voie CG (CLIC/GEEC) sont principalement des protéines avec des ancres GPI mais aussi des protéines transmembranaires telles que CD44 (Howes et al. 2010).

Il existe aussi l’endocytose ARF6 (ADP ribosylation factor 6) qui est clathrine et dynamine indépendante (Figure 4). Elle possède beaucoup de similitudes avec la voie CG à la différence qu’ici le rôle de ARF6 est essentiel (Radhakrishna et Donaldson 1997). Comme pour la voie

CG, les cargos internalisés par la voie ARF6 sont des protéines possédant une ancre GPI mais aussi des protéines transmembranaires telles que CD44, GLUT1 (Glucose Transporter 1),

CD98, Basigin (CD147) ou encore des protéines de l’immunité comme CD1a (Eyster et al.

2009).

Il existe donc de nombreuses catégories d’endocytose faisant intervenir différents complexes protéiques régulant la formation des vésicules d’endocytose, leur scission et leur transport. Chacune de ces voies semble associée à la sélection de cargos spécifiques même si certains sont communs à plusieurs voies. L’un des points communs à toute endocytose est le rôle essentiel dans la courbure de la membrane plasmique et les tensions générées induites par la polymérisation de l’actine. L’autre point commun est que l’ensemble des vésicules d’endocytose fusionnent avec les endosomes précoces.

23 24

1.2.1.4 Endosomes précoces : gares de triage

Les endosomes précoces sont des structures membranaires composées essentiellement de PI(3)P, mais la présence de PI(4,5)P2, PI(3,4)P2 et de PI(4)P y est aussi détectée (Vicinanza et al. 2008). L’ensemble du contenu vésiculaire issu de l’endocytose converge vers les endosomes précoces et fusionne avec ceux-ci (Eisenberg-Bord et al. 2016).

Avant de fusionner, les vésicules d’endocytoses doivent se maintenir à proximité l’une de l’autre afin de stabiliser les complexes de fusion des membranes (Cruz et Kim 2019). EEA1

(Early Endosome Antigen 1) est l’une des principales protéines d’ancrage médiant la fusion des vésicules d’endocytose (Christoforidis et al. 1999). La présence de SNARE à la surface des endosomes régule la fusion des vésicules d’endocytose avec les endosomes précoces. En effet, la présence de quatre SNARE, soit Qa, Qb, Qc et R-SNARE permet la formation d’un complexe stable qui induit la fusion de deux membranes lipidiques (Südhof et Rothman

2009). Il a aussi été montré que les endosomes précoces peuvent fusionner entre eux, fusion dite homotypique, ou avec des endosomes de recyclage (ER), fusion dite hétérotypique (Cruz et Kim 2019). Chacun de ces évènements est médié par l’action de SNARE telles que la

STX13 (Syntaxin 13) ou VAMP4 (Vesicle Associated Membrane Protein 4) (Skalski et al.

2010). Aussi, il a été montré que les dynéines, protéines motrices, permettent le transport des endosomes précoces le long des microtubules vers des régions périnucléaires, appelées centres des endosomes de recyclage (Driskell et al. 2007). Ce mécanisme induit la transition des endosomes précoces en endosomes de recyclage. De la même manière, les endosomes précoces peuvent évoluer en endosomes tardifs et enfin fusionner avec les lysosomes

(Huotari et Helenius 2011). Les endosomes précoces sont donc des organites extrêmement dynamiques et représentent le nerf du triage endosomal (Figure 5).

24 25

Figure 5 : L’endosome précoce, centrale de triage L’endosome précoce représente le point névralgique de la voie endosomale. Les cargos endocytés vont transiter par l’endosome précoce ou ils seront redirigés vers le recyclage ou la dégradation. Les cargos recyclés sont triés via la formation des tubules endosomaux qui émergent de manière spontanée ou de façon régulée par les complexes WASH ou Retromère. Les cargos sont triés soit directement à la membrane plasmique soit au Golgi ou aux endosomes de recyclage. EHD1 permet la sélection des cargos pour le triage vers les endosomes de recyclages. Le complexe ESCRT, contenant la protéine HRS, permet l’internalisation de cargos au sein de microvésicules et ainsi la maturation de l’endosome précoce en endosome tardif. Le complexe HOPS permet la fusion de l’endosome tardif avec les lysosomes et donc la dégradation de son contenu.

25 26

En effet, de nombreux complexes protéiques permettent de moduler la morphologie de l’endosome précoce et génèrent des microdomaines nécessaires au triage de cargos spécifiques (van Weering et al. 2012). Ainsi, c’est aux endosomes précoces qu’est décidé le destin des cargos endocytés. Les cargos transitant par les endosomes précoces peuvent être triés soit vers la dégradation lysosomale, soit recyclés vers le réseau transgolgien via le transport rétrograde ou encore recyclés à la membrane plasmique (Figure 5). Chaque voie de triage mobilise des complexes protéiques spécifiques, et possède ses particularités.

1.2.1.5 Dégradation lysosomale

Tout d’abord, le triage vers la dégradation lysosomale s’effectue par la reconnaissance de cargos ubiquitinylés. Les motifs ubiquitinylés présents sur la partie cytosolique des cargos endocytés sont reconnus par le motif UIM (Ubiquitin Interacting Motif) de la protéine HRS

(Hepatocyte growth factor-Regulated tyrosine kinase Substrate) (Shih et al. 2002). HRS interagit aussi avec la clathrine formant ainsi des microdomaines spécifiques. Enfin, HRS recrute les membres du complexe ESCRT-I (Endosomal Sorting Complex Required for

Transport) lequel recrute successivement ESCRT-II et -III. Les complexes ESCRT ont pour effet de générer des invaginations membranaires formant ce qu’on appelle des corps multivésiculaires (MVB; Multivesivular Body) (Williams et Urbé 2007). La maturation des

MVB conduit au recrutement du complexe HOPS (Homotypic Fusion and Protein Sorting) qui comporte des SNARE permettant de médier la fusion avec les lysosomes (Spang 2016).

Enfin, les cargos sont dégradés par les enzymes lysosomales (Piper et Katzmann 2007). Le transport rétrograde des endosomes précoces vers le réseau transgolgien permet de connecter la voie d’endocytose avec la voie de sécrétion et de réinjecter les cargos endocytés dans la voie de sécrétion (Figure 5).

26 27

1.2.1.6 Recyclage endosomal

Les cargos endocytés, en fonction des besoins de la cellule, peuvent être recyclés directement à la membrane plasmique. On dissocie deux types de recyclage endosomal. Un recyclage, dit lent, qui nécessite le passage des cargos via les endosomes de recyclage ou le recyclage, dit rapide, qui permet un retour direct des cargos de l’endosome précoce à la membrane plasmique via la formation de tubules de recyclage (Naslavsky et Caplan 2018).

Le recyclage de type lent correspond au transit de cargos via les endosomes de recyclage

(Figure 5). Les endosomes de recyclage sont des structures tubulaires et vésiculaires qui s’accumulent dans des régions périnucléaires au niveau du centre d’organisation des microtubules (Taguchi 2013). Les cargos sont majoritairement issus des endosomes précoces mais des évidences montrent qu’il existe des échanges directs entre le Golgi, la membrane plasmique et les endosomes de recyclage. La famille des protéines EHD (EH Domain

Containing) est fortement liée aux endosomes de recyclage. Les protéines EHD sont caractérisées par la présence d’un domaine EH, en C-terminal, impliqué dans des interactions protéine-protéine (Naslavsky et Caplan 2011). La protéine EHD1 est considérée comme des ciseaux qui permettent la fission de tubules ou vésicules formées aux endosomes précoces.

De plus, EHD1 interagit avec les protéines MICAL-L1 (Microtubule Associated

Monooxygenase, Calponin And LIM Domain Containing 1-Like 1) et CRMP2 (Collapsin response mediator protein-2) régulant le transport, des endosomes précoces vers les endosomes de recyclages, de vésicules le long des microtubules et ce, via le contrôle des dynéines (Naslavsky et Caplan 2018). Aussi, EHD3 permet de stabiliser les tubules de recyclages au niveau des endosomes de recyclage en interagissant avec MICAL-L1, protéine

27 28 qui possède un domaine BAR pouvant induire des courbures de la membrane (Bahl et al.

2016).

Le recyclage de cargos des endosomes précoces à la membrane plasmique nécessite deux évènements essentiels : la reconnaissance du cargo et la formation de tubules de recyclage qui permettent d’isoler les cargos recyclés au sein de structures spécifiques du reste de l’endosome précoce (Van Weering et al. 2012). De nombreux complexes et protéines peuvent sélectionner des cargos et induire leur recyclage rapide. Les SNX ont la capacité d’être recrutées aux endosomes précoces via l’interaction entre leur domaine PX et le PI(3)P composant la membrane des endosomes précoces (Van Weering, et al. 2012). En revanche, les SNX ont aussi la capacité d’interagir directement avec certains cargos. Ainsi, SNX17 est impliquée dans le recyclage de l’intégrine 5 et autres protéines avec un domaine FERM

(Steinberg et al. 2012). De plus, SNX17 interagit avec le domaine cytosolique de LRP1

(Lipoprotein Receptor-Related Protein 1) (Donoso et al. 2009). D’autres protéines que les

SNX sont aussi impliquées dans des évènements de recyclage. Par exemple, la protéine

ACAP1 (ArfGAP With Coiled-Coil, Ankyrin Repeat And PH Domains 1) interagit avec les domaines cytosoliques du récepteur à la transferrine (TfR), de GLUT4, et des intégrines régulant leur triage vers la membrane plasmique (J. Li et al. 2007). Enfin, les trois complexes

CCC, Retriever et Retromer agissent en concert avec les SNX pour réguler le recyclage de cargos spécifiques (McNally et Cullen 2018). En effet, le Retriever, composé des sous unités

VPS29, VPS35L et VPS26C interagit avec SNX17 et régule le recyclage de l’intégrine 1.

Aussi le complexe CCC interagit avec la sous unité VPS35L du Retriever et régule le recyclage de récepteurs aux lipoprotéines à faible densité (McNally et al. 2017).

28 29

L’un des premiers complexes ayant été caractérisé dans le triage endosomal est le Rétromère.

Par ailleurs, le complexe du Rétromère a été caractérisé pour son rôle dans le trafic rétrograde de CI-MPR (Cation-independent Mannose 6-Phosphate Receptor), impliqué dans le transport d’enzymes glycosylées du Golgi aux lysosomes, des endosomes précoces vers le réseau transgolgien, permettant ainsi la connexion entre le système d’endocytose et celui de la sécrétion (Seaman 2004).

1.2.2 Le complexe Rétromère

Découvert chez la levure sous la forme d’un heptamère, le Rétromère est majoritairement localisé aux endosomes précoces et associé à des évènements de triage. En effet, le Rétromère interagit avec le domaine cytosolique d’une large gamme de protéines liées à la membrane des endosomes précoces. Le recrutement du Rétromère est un signal pour la formation de tubules et le recyclage de cargos spécifiques (Harbour et al. 2010).

On peut cependant distinguer deux sous-complexes. Un premier hétérotrimère composé des sous-unités VPS35 (Vacuolar Protein Sorting-Associated Protein 35), VPS29 et VPS26A ou B formant un complexe de 150 kDa (Figure 6). Ce trimère est considéré comme le cœur du Rétromère, il sera par la suite nommé CSC (Cargo Selective Complex)

(Hierro et al. 2007). Les sous-unités VPS29 et VPS26 interagissent avec les domaines C- terminal et N-terminal de VPS35, respectivement. Les sous unités VPS26 A et B sont deux paralogues permettant la formation de deux complexes du Rétromère avec des fonctions distinctes (Kerr et al. 2005). En effet, le Rétromère a été initialement caractérisé comme étant impliqué dans le trafic rétrograde de CI-MPR de l’endosome précoce vers le Golgi. Or, le complexe du Rétromère formé avec la sous-unité VPS26B est incapable de lier le CI-MPR

(Bugarcic et al. 2011). La surexpression de VPS26B inhibe l’expression de VPS26A et induit

29 30 la dégradation lysosomale de CI-MPR. De plus, les complexes formés par VPS26A ou B arborent des localisations sensiblement différentes puisque VPS26B est associé à des endosomes plus avancés dans leur maturation comparativement à VPS26A (Bugarcic et al.

2011).

Le trimère formé par VPS35-VPS29-VPS26 est associé à un dimère composé des

SNX1/2 et SNX5/6. Il existe quatre possibilités de dimère, SNX1 peut être associé soit à

SNX5 soit à SNX6. De même, SNX2 peut être associé à SNX5 ou SNX6. Les SNX1,2,5 et

6 sont des protéines possédant un domaine PX et un domaine BAR (Wassmer et al. 2009).

Le domaine PX permet l’association avec la membrane de l’endosome précoce qui est riche en PI(3)P (Ellson et al. 2002). Le domaine BAR permet d’induire des courbures de la membrane et de participer à la formation de tubules. L’interaction entre le CSC et les dimères

SNX1/2 et SNX5/6 a été mis en évidence par des expériences d’imagerie confocale et de co- immunoprécipitation (Kurten et al. 2001). Cependant, des données récentes d’imagerie à haute résolution suggèrent que ces deux sous complexes occupent des microdomaines distincts sur l’endosome. Aussi, des études de déplétion des membres du Rétromère par l’approche CRISPR/Cas9 suggèrent que le CSC n’est pas impliqué dans le transport rétrograde de CI-MPR (Kvainickas et al. 2017). D’autres données montrent que le trimère

VPS35, 29 et 26 est bien impliqué dans le transport rétrograde de CI-MPR à la différence qu’ici le CSC interagit avec la golgine-GCC88 (Cui et al. 2019). Par ailleurs, le transport rétrograde du CI-MPR régulé par les SNX1 et 5 nécessite la golgine-245 (Shin et al. 2017).

Ainsi, les dimères SNX1/2 et SNX5/6 semblent avoir des fonctions indépendantes du CSC.

Enfin, des données montrent que le CSC est aussi associé aux SNX3 et SNX27 qui ne possèdent pas de domaine BAR (Temkin et al. 2011; Harterink et al. 2010). Par ailleurs, l’association du CSC avec les SNX3 et 27 semblent plus stable que celle avec les dimères

30 31

SNX1/2 et SNX5/6. Ainsi, le Rétromère apparait comme un complexe versatile dont la constance est maintenue par la présence du CSC composé des sous-unités VPS35, VPS29 et

VPS26, lesquelles sont associées à des SNX (Figure 6). Enfin, le Rétromère est principalement recruté aux endosomes via SNX3 ou RAB7 (Harrison et al. 2014).

Les dernières études de cristallographie ont permis de mettre en évidence les manteaux membranaires formés par le CSC en association avec SNX1, SNX3 et SNX27

(Kovtun et al. 2018; Lucas et al. 2016). À la différence des manteaux de clathrine, COPI et

COPII qui forment des sphères, le manteau formé par le Rétromère et SNX1 arbore une structure symétrique en hélice (Figure 6). On distingue deux niveaux avec une couche inférieure composée de la juxtaposition de SNX1 qui courbe la membrane, et le trimère du

CSC qui forme la couche extérieure du manteau. Les données montrent que le CSC est relié

à SNX1 par la sous unité VPS26, et que VPS35 et VPS29 forment un V où VPS35 représente la base et VPS29 le sommet. Dans cette structure, VPS29 est orienté vers le cytosol, et seul

SNX1 semble être en contact avec la membrane de l’endosome (Lucas et al. 2016). De plus,

SNX1 forme des dimères induisant des pseudo boucles sur de courtes distances permettant une plasticité dans la courbure de la membrane. Du fait de l’hétérogénéité de l’ancrage de

SNX1 à la membrane, la sous-unité VPS26 peut s’ancrer dans différentes orientations respectant des espaces plus ou moins grands. La présence d’espace et la variabilité dans l’orientation du CSC permet donc l’interaction de VPS29 avec différents partenaires protéiques.

De manière différente, lorsque le CSC est associé à SNX3, il forme des structures en

T. SNX3 interagit avec le domaine C-terminal de VPS26 et N-terminal VPS35 au sommet du T. Ici, SNX3 et VPS26 sont tous les deux en contact avec la membrane plasmique (Figure

6).

31 32

Figure 6 : Structures cristallographiques des différents manteaux formés par le Rétromère (A) Structure en forme de T formée par le Rétromère lorsque celui-ci est associé à SNX3. VPS26 et VPS35 interagissent avec SNX3, et seules VPS26 et SNX3 sont en contact avec la membrane. (B) Structure en forme de T formée par le Retromère en association avec SNX27. De la même manière que pour SNX3, seules SNX27 et VPS26 sont en contact avec la membrane. (C) Modèle de l’organisation du Rétromère en association avec les SNX possédant un domaine BAR. Ici, le CSC n’est plus en contact avec la membrane et est orienté vers le cytosol. Ce sont les SNX qui permettent l’ancrage à la membrane. (D) Modèle d’organisation des complexes de Rétromère pour former un manteau et induire des courbures de la membrane. (Figure tirée de Lucas et al. 2016, autorisation d’emprunt de la figure obtenue de Elsevier)

32 33

Il a été montré que l’interface entre VPS26 et SNX3 forme une poche permettant la reconnaissance de cargos tels que DMT1-II (Lucas et al. 2016). De même, lorsque le CSC est associé à SNX27, il arbore une structure similaire à celle du CSC avec SNX3, en forme de T, où le domaine PDZ de SNX27 interagit avec VPS26. Ici VPS26 est aussi en contact avec la membrane de l’endosome et l’interaction de VPS26 avec SNX27 stabilise fortement l’interaction du domaine PDZ avec les différents cargos. SNX3 et 27 ne possédant pas de domaine BAR, c’est le recrutement d’effecteurs tels que MON2, DOP1B et ATP9A par le

CSC qui induit les courbures de la membrane et la formation de tubules (McGough et al.

2018). Longtemps le trimère VPS35, 29 et 26 a été considéré comme le sous-complexe régulant la sélection du cargo. Il semble que la réalité soit plus complexe et que les SNX associées au CSC jouent aussi un rôle important dans la sélection des cargos.

Le CSC joue un rôle dans la sélection des cargos triés par interactions directes avec ceux-ci. En effet, plusieurs études montrent que VPS26 et VPS35 interagissent directement avec le domaine cytosolique des cargos endosomals (Seaman 2012). La plupart des cargos reconnus par le Rétromère possède un motif aromatique et hydrophobique F/W-L-M/V

(McNally et Cullen 2018). Longtemps VPS35 a été considéré comme la seule plateforme de sélection des cargos. Cependant, des études montrent que VPS26 permet d’induire le recyclage de Sorl1 par interaction avec son domaine cytosolique via la reconnaissance du motif FANSHY (Phe-Ala-Asn-Ser-His-Tyr) (Fjorback et al. 2012).

Du fait de sa versatilité, le Rétromère est impliqué dans le triage d’un grand nombre de cargos. Par exemple, lorsqu’il est associé à SNX3, le Rétromère est impliqué dans les transports rétrogrades de Wls, un récepteur à WNT, ainsi que du CI-MPR (Belenkaya et al.

2008). Ce même complexe est impliqué dans le transport rétrograde du récepteur de DMT1-

II (Lucas et al. 2016). Aussi, des données montrent que SNX12 et SNX3 possèdent des

33 34 fonctions similaires et sont tous deux impliqués dans le recyclage du récepteur à la transferrine (Priya et al. 2017).

En association avec SNX27, le Rétromère régule principalement le recyclage de cargos de l’endosome à la membrane plasmique. Le domaine PDZ de SNX27 a été caractérisé pour interagir avec des centaines de protéines transmembranaires tels que des GPCR ou des canaux ioniques (Gallon et al. 2014). Aussi, il a été montré que le Rétromère lié à SNX27 permet le recyclage à la membrane plasmique du récepteur  adrénergique de type II (2AR) via la reconnaissance du domaine cytosolique FERM de 2AR (Temkin et al. 2011). Par ailleurs, l’association de CSC avec SNX27 induit le recyclage à la membrane plasmique de

GLUT1 (Kvainickas et al. 2017). En effet, la déplétion de VPS35 inhibe le recyclage à la membrane plasmique du récepteur GLUT1, limitant l’import de glucose. Enfin, il a été montré que le domaine PDZ de SNX27 permet le recrutement et le recyclage à la membrane plasmique de cargos phosphorylés tels que le NMDA (Clairfeuille et al. 2016).

Lorsqu’il est associé au dimère SNX1/2 SNX5/6, le Rétromère régule le transport rétrograde de cargos tels que le CI-MPR et joue aussi un rôle important dans l’inhibition de la voie de dégradation. En effet, le Rétromère permet le recrutement de RME-8 (Receptor mediated endocytosis 8) et de TBC1D5 (TBC1 domain family member 5) qui inhibent respectivement le complexe ESCRT et RAB7 tous deux impliqués dans la voie de dégradation lysosomale (Norris et al. 2017). Par ailleurs, l’inhibition de RAB7 par le

Rétromère joue un rôle important dans la régulation de la mitophagie et de la signalisation mTORC1 au lysosome (Kvainickas et al. 2019). Ainsi, le Rétromère est impliqué dans le devenir d’un grand nombre de cargos et joue un rôle important dans l’homéostasie cellulaire et d’autant plus dans les cellules à forte activité sécrétrice. En effet, des mutations dans les

34 35 sous-unités du Rétromère sont impliquées dans de nombreuses pathologies neurodégénératives telles que la maladie de parkinson ou d’Alzheimer (Brodin et Shupliakov

2018).

Le Rétromère n’est pas seulement impliqué dans la sélection de cargos. En effet, le sous complexe CSC, initialement identifié comme le cœur de sélection des cargos, régule l’ensemble des évènements de triage via le recrutement d’une large gamme d’effecteurs.

Comme il a été mentionné, le CSC permet la sélection de cargos et l’induction de courbures de la membrane via l’interaction avec les SNX à domaine BAR ou la protéine MON2

(McGough et al. 2018). Le CSC permet aussi de délimiter les domaines de maturation des endosomes par l’inhibition de RAB7 ou ESCRT (Norris et al. 2017). Enfin, le CSC via l’interaction de VPS35 avec FAM21 permet le recrutement aux endosomes du complexe

WASH (Wiskott Aldrich Syndrome protéine et SCAR Holomogue) qui est impliqué dans la régulation de la polymérisation de l’actine (Gomez et Billadeau 2009). De la même manière que pour les évènements d’endocytose, la polymérisation de l’actine aux endosomes joue un rôle important dans la régulation des évènements de triage.

1.2.3 Le complexe WASH

La polymérisation d’actine, c’est-à-dire l’incorporation de monomères d’actine dans des filaments (F-actine), joue des rôles importants dans la migration, l’adhésion, la cytokinèse et le transport membranaire (Simonetti et Cullen 2019; Kovacs et al. 2011). Le complexe ARP2/3 régule la polymérisation d’actine aux différents compartiments cellulaires

(Rotty et al. 2013). ARP2/3 est recruté à chaque organite par différents complexes comportant une protéine de la famille WASP (Kollmar et al. 2012). La protéine WASP a été identifiée suite à l’identification d’une mutation présente chez les patients atteints de la

35 36 maladie de l’X fragile ou syndrome de Wiskott–Aldrich. Aux endosomes, c’est le complexe

WASH qui assure le recrutement et l’activation du complexe ARP2/3 (Duleh et Welch 2010).

Le complexe WASH a une structure similaire au complexe WAVE qui induit la polymérisation d’actine au front de migration (Jia et al. 2010). En effet, ce complexe comporte la protéine WASH1 (WASP and SCAR homolog) qui appartient à la famille

WASP.

Le complexe WASH est conservé chez l’ensemble des eucaryotes. Des études montrent que l’inhibition de WASH provoque des défauts de développement de la larve chez la drosophile (Liu et al. 2009). En revanche, d’autres évidences suggèrent que différents mutants de WASH n’affectent pas la viabilité de la drosophile (Nagel et al. 2017). Cependant la délétion de WASH chez la souris est létale au stade embryonnaire (Xia, Wang, Huang, Du, et al. 2014). Le complexe WASH est un pentamère (Figure 7) composé des sous-unités

WASH1 (WASHC1), FAM21A/C (WASHC2A/C), CCDC53 (WASHC3), SWIP

(WASHC4), et Strumpellin (WASHC5) (Jia et al. 2010). La perte d’un de ces membres a pour effet d’affecter la stabilité des autres sous-unités. Par exemple, la déplétion de WASH1 affecte les niveaux d’expression de CCDC53 dans des cellules MEF (Mouse Embryonic

Fibroblast) (Jia et al. 2010). WASH1 possède un domaine VCA (Verpolin Connecting and

Acidic) qui recrute ARP2/3, et une activité NPF (Nucleation-Promoting Factor) qui induit l’activité d’ARP2/3 (Xia et al. 2013). Aussi, WASH1 présente un domaine WHD1 (WASH homology domain 1) et un domaine WHD2 (WASH homology domain 2) qui permettent l’interaction avec la tubuline (Gomez and Billadeau 2009). FAM21A/C est constitué de deux régions : une « tête » et une « queue » (Figure 7). La tête de FAM21 sert de ciment permettant l’assemblage du complexe WASH. La queue de FAM21 est un enchainement de 21 motifs

LFa qui interagit avec VPS35 (D. Jia et al. 2012). De plus, la queue de FAM21 est une

36 37 plateforme qui permet l’interaction avec de nombreuses protéines telles que CapZ,

ANKRD50 (Ankyrin Repeat Domain 50), FKBP15 (FK506-Binding Protein 15), TBC1D23, le complexe CCC, SNX1 et RME-8 (Wang et al. 2018).

Figure 7 : Relation des complexes WASH et Rétromère Le complexe WASH formé des sous-unités Strumpellin, KIAA1033, WASH1, CCDC53 et FAM21 est recruté au complexe CSC du Rétromère, composé de VPS35, VPS26, et VPS29, via l’interaction de VPS35 avec la queue de FAM21. L’ubiquitinylation de WASH1 par TRIM27 permet le recrutement de ARP2/3 lequel induit la polymérisation de l’actine. La déubiquitinylation de WASH1 par USP7 permet un cycle d’activation et d’inhibition de WASH1. Le complexe WASH peut aussi être recruté par HRS. La polymérisation d’actine permet alors de délimiter l’induction de vésicules intraluminales induites par le complexe ESCRT. Le complexe du Rétromère peut être soit recruté par les SNX3 et 27 ou par RAB7. La sous-unité VPS29 du Rétromère peut interagir avec TBC1D5, une GAP qui inhibe RAB7.

Le complexe WASH est principalement localisé aux endosomes. Ainsi son recrutement est médié par son interaction avec le Rétromère et sa capacité à interagir avec des lipides chargés négativement (Helfer et al. 2013). Cependant, la déplétion de FAM21 qui interagit avec le Rétromère réduit la présence de WASH aux endosomes précoces mais ne l’inhibe pas complètement. Cela suggère qu’il existe d’autres mécanismes, indépendants du

Rétromère, de recrutement de WASH. Ainsi, de nouvelles données montrent que le complexe

37 38

WASH est recruté aux endosomes via HRS, qui est un membre du complexe de dégradation lysosomale (Figure 7). Par ailleurs, WASH interagit avec SNX27, le complexe CCC, et

SNX1 via RME-8 qui pourraient induire son recrutement aux endosomes précoce.

L’activité du complexe WASH est régulé par la poly-ubiquitinylation de la lysine 220 de WASH1 par l’ubiquitine ligase TRIM27 qui entraine une augmentation de l’activité de polymérisation d’actine (Hao et al. 2013). En effet, l’ubiquitination de WASH1 perturbe une région inhibitrice du domaine VCA. TRIM27 est recruté aux endosomes via l’interaction avec MAGE-L2 (MAGE-like protein 2) qui lui-même interagit avec le Rétromère. Par ailleurs, USP7 (Ubiquitin carboxyl-terminal hydrolase 7) interagit aussi avec MAGE-L2 et permet la déubiquitinylation de la lysine 220 de WASH1 et l’inhibition de son activité.

TRIM27 et MAGE-L2 ne sont pas exprimées de manière ubiquitaire; il doit donc exister d’autres mécanismes de régulation du complexe WASH. Ainsi, il a été montré, in vivo que

Rho1 est impliqué dans l’activation de WASH mais cet effet n’a pu être confirmé in vitro

(Liu et al. 2009).

Dans un premier temps, la polymérisation d’actine aux endosomes précoces médiée par WASH permet de générer des frontières entre les domaines destinés au recyclage et ceux dirigés vers la dégradation lysosomale (MacDonald et al. 2018). En effet, la polymérisation d’actine aux endosomes permet d’inhiber la diffusion latérale et passive de cargos à travers la membrane de l’endosome. Cela a pour effet de séquestrer les cargos dans des microdomaines et d’inhiber leur internalisation dans des vésicules intraluminales et leur dégradation lysosomale. Le fait que HRS puisse recruter WASH suggère qu’il existe une réelle coordination entre le triage vers la dégradation et le recyclage (Figure 7). Il a été montré qu’il existe des motifs de liaison à l’actine dans le domaine cytosolique de certains cargos.

38 39

L’interaction des cargos avec l’actine les oriente vers le recyclage. Il existe donc un équilibre entre la dégradation induite par l’ubiquitinylation de cargo qui est reconnue par le complexe

ESCRT et le recyclage favorisé par la liaison à l’actine induite par WASH. En effet, l’ajout de domaine de liaison à l’actine dans la partie cytosolique de 2AR est suffisant pour induire son recyclage à la membrane plasmique et d’inhiber sa dégradation lysosomale, respectivement (Puthenveedu et al. 2010). De plus, la V-ATPase interagit avec une forte affinité avec la F-actine et cette interaction est suffisante pour le recyclage de la V-ATPase

(Vitavska et 2005). Ainsi, il semble que les cargos interagissent directement avec l’actine branchée et cela influe directement leur triage endosomal vers le recyclage. D’autre part, la formation des corps multivésiculaires à l’origine de la maturation des endosomes est induite et délimitée par la polymérisation de l’actine régulée par l’annexine A2 (Mayran et al. 2003).

Cela a pour effet d’ajouter un niveau supplémentaire de ségrégation des domaines de recyclage et de dégradation.

Ensuite, le complexe WASH participe à la formation de sous-domaines de recyclage en concentrant la présence de cargos via l’interaction de plusieurs complexes du Rétromère avec la « queue » de FAM21, et par interaction directe des cargos avec l’actine (Simonetti et

Cullen 2019). L’actine joue aussi un rôle dans l’architecture de ces sous-domaines endosomaux puisque la déplétion de ARP2/3 induit la fusion des domaines WASH et

Rétromère (Derivery et al. 2012). La polymérisation d’actine régulée par WASH est donc impliquée dans le transport rétrograde de CI-MPR médiée par le Rétromère, SNX3, et les

SNX1/2 et SNX5/6. WASH est aussi impliqué dans le recyclage à la membrane plasmique du TfR, du 2AR, de GLUT1, du transporteur de cuivre ATP7A et des intégrines 51

(Phillips-Krawczak et al. 2015). Leur recyclage est dépendant du Rétromère, excepté celui

39 40 des intégrines 1 qui est associé au Retriever (McNally et al. 2017). De manière étonnante,

WASH a aussi été montré pour jouer un rôle dans la régulation de la dégradation lysosomale du récepteur à l’EGF (Epiderrmal Growth Factor). Ce qui suggère que le rôle de WASH est complexe et qu’il est influencé par ses partenaires protéiques. Ici, SNX6 semble requise pour le triage de l’EGFR (Epiderrmal Growth Factor Receptor) vers la dégradation (Cavet et al.

2008).

WASH régule la formation des tubules de recyclage aux endosomes. D’abord, les complexes de sélection des cargos induisent des torsions de la membrane via la présence de

SNX avec un domaine BAR. En synergie, WASH est recruté par interaction des complexes de sélections avec FAM21. Ensuite, WASH recrute et active ARP2/3 qui induit la polymérisation d’actine (Simonetti et Cullen 2019). La polymérisation d’actine, et la présence de protéines motrices favorisent la formation des tubules induites par les SNX.

Enfin, les tensions induites par la polymérisation d’actine favorisent la fission de tubules de recyclage (Simonetti et Cullen 2019). De plus, les protéines motrices associées aux microtubules sont recrutées par WASH et jouent un rôle dans la fission des tubules (Derivery et al. 2009). Cet effet, a été observé pour des vésicules contenant la métalloprotéase MT1-

MMP (Marchesin et al. 2015). Enfin, il a été montré que les points de contact membranaire entre les endosomes et le RE permettent la scission des tubules (Rowland et al. 2014). Des données montrent que ces points de contact sont localisés sur des domaines endosomals contenant FAM21. Ainsi, il a été montré que la déplétion de WASH induit l’accumulation de longs tubules émergeant de l’endosome précoce et perturbe le transport rétrograde de CI-

MPR (Gomez et Billadeau 2009).

40 41

Deux types de tubules de recyclage aux endosomes précoces ont été caractérisés : ceux formés spontanément et ceux régulés par la polymérisation d’actine (Maxfield et

McGraw 2004; Simonetti et Cullen 2019). WASH est impliqué dans la formation des tubules présentant des filaments branchés d’actine (Gautreau et al. 2014). Les cargos tels que 2AR sont recyclés à la membrane plasmique via des tubules de recyclage qui sont caractérisés par la présence de ARP2/3 et du complexe WASH (Simonetti et Cullen 2019). Par ailleurs, le transport rétrograde de Wls induit par le Rétromère dépend aussi de tubules riches en actine

(Belenkaya et al. 2008). La composition en cargos de ces tubules est distincte de ceux formés spontanément. Ainsi, la stabilisation de ces tubules par la présence d’actine permet l’accumulation de cargos qui diffusent lentement à travers la membrane de l’endosome

(Puthenveedu et al. 2010). Certains récepteurs de la classe des GPCR ou RTK sont actifs au niveau de l’endosome (Sorkin et Von Zastrow 2009). Ce sont des récepteurs qui ont un temps de résidence considéré comme long aux endosomes. De plus, il a été constaté que les GPCR actifs aux endosomes, le sont aux niveaux de plateformes riches en actine (Pavlos et

Friedman 2017). Enfin, contrairement aux récepteurs transitant via les tubules formés spontanément, les cargos accumulés dans les tubules de recyclage possèdent une séquence de triage reconnue par les complexes de sélection tels que le Rétromère, le Retriever, le CCC ou par l’actine (Simonetti et Cullen 2019). On est donc en présence d’un recyclage à deux vitesses avec un triage spontané et un triage impliquant de nombreux mécanismes moléculaires et étant associé à des évènements de signalisation cellulaire. La polymérisation d’actine induite par le complexe WASH permet donc de générer une plateforme et de réguler de manière active le recyclage vers l’ensemble vers la membrane plasmique et le Golgi. Le complexe WASH ne régule pas seulement la formation et la composition de tubules de

41 42 recyclages mais assure le maintien de l’architecture des compartiments composant le trafic endolysosomal (Gomez et al. 2012). En effet, la déplétion de WASH dans des MEF induit la relocalisation et la concentration des endosomes et lysosomes dans des régions périnucléaires.

La polymérisation d’actine induite par WASH est impliquée dans de nombreux processus indépendants de la formation de tubules de recyclage. En effet, WASH assure le maintien de l’architecture du noyau (Verboon et al. 2015). WASH interagit avec les lamines et permet la polymérisation d’actine dans le noyau. Par ailleurs, WASH interagit avec le complexe NURF (Nucleosome-Remodeling Factor) et permet son recrutement au promoteur de Myc (Xia et al. 2014). De plus, WASH1, Strumpellin et SWIP sont des composants du complexe de transcription TRF2 (Telomeric Repeat-binding Factor 2). Tous les membres du complexe WASH excepté CCDC53, possèdent une séquence NLS (Nuclear Factor

Sequence) (Alekhina et al. 2017). Enfin, FAM21 est impliquée dans la transcription de NF-

B (Deng et al. 2015). Le complexe WASH est aussi impliqué dans la régulation de l’autophagie. La déplétion de WASH chez la souris induit une inhibition du flux autophagique menant à la mort de l’embryon. En effet, WASH recrute l’ubiquitine ligase E3

RFN2 qui ubiquitine AMBRA1 (Activating molecule in BECN1-Regulated Autophagy protein 1), un activateur de VPS34 et un inhibiteur de mTOR, et induit sa dégradation et inhibant donc l’activation de beclin-1 et VPS34 (Xia et al. 2014). Une autre étude montre que l’inhibition de l’expression de WASH dans des neuroblastomes affecte le trafic membranaire d’ATG9A et inhibe le flux autophagique (Zavodszky et al. 2014).

De la même manière que le Rétromère, des défauts d’activité de WASH sont impliquées dans de nombreuses maladies neurodégénératives. La mutation de VPS35

42 43

(D620N) associée avec la maladie de Parkinson inhibe l’interaction de WASH1 avec le

Rétromère et affecte l’autophagie et le transport vésiculaire issu des endosomes tardifs

(Zavodszky et al. 2014). Des mutations gain de fonction dans RME-8 qui interagit avec

FAM21 sont identifiées chez des patients avec la maladie de Parkinson (Seaman et Freeman

2014). Enfin, des mutations dans les membres du complexe WASH sont impliquées dans de nombreuses pathologies neurologiques. La mutation du gène SWIP (P1019) et de Strumpellin induisent la diminution de l’expression des autres membres du complexe WASH et sont associées avec des désordres intellectuels (Jia et al. 2010). Des mutations de Strumpellin sont

également présentes chez des patients atteints de paraplégie spastique héréditaire (HSP).

1.3 Les régulateurs du trafic membranaire

1.3.1 Les RAB

1.3.1.1 Généralités

L’ensemble du trafic membranaire est donc extrêmement complexe et fait intervenir un grand nombre de complexes protéiques. Pour exemple, le recyclage de cargos à la membrane plasmique depuis l’endosome précoce, via les tubules de recyclage, nécessite : 1)

La définition de domaines de recyclage induits par le complexe WASH, 2) la reconnaissance des cargos par des complexes spécifiques tels que le Rétromère, 3) l’induction de courbures de la membrane par les SNX, 5) la fission du tubule médiée par la polymérisation d’actine via APR2/3, 6) le transport des tubules via des protéines motrices et enfin leur fusion avec la membrane plasmique régulée par les SNARE. De plus, il existe plusieurs complexes de sélection des cargos qui reconnaissent différents types de protéines transmembranaires.

Aussi, un même cargo peut être reconnu par différents complexes. L’ensemble de ces acteurs

43 44 du trafic membranaire agissent de manière successive et/ou concomitante, ils doivent donc

être coordonnés afin d’assurer l’homéostasie du trafic membranaire. Les RAB (Touchot et al.1987) sont considérées comme les chefs d’orchestre qui permettent l’harmonisation de ce trafic (Pfeffer et Kellogg 2017). En effet, les RAB sont des plateformes qui recrutent des effecteurs régulant l’ensemble des grandes étapes du trafic membranaire (Gillingham et al.

2014). De plus, les RAB permettent de maintenir l’intégrité de l’identité membranaire de chaque organite (Jean et Kiger 2012).

1.3.1.2 Structure des RAB

Les RAB sont des protéines monomériques de 21 à 25 kDa (Dumas et al. 1999), elles sont extrêmement conservées au cours de l’évolution de la levure à l’homme (Brighouse et

Field 2010). Les RAB sont de petites GTPases qui alternent entre une forme liée au GTP et une forme liée au GDP. Avec environ 70 membres, les RAB représentent la plus grande famille de GTPase (Barr et Lambright 2010). Elles sont impliquées dans la régulation du trafic membranaire mais aussi dans la croissance, la migration, la division cellulaire et la transduction de signaux (Hutagalung et Novick 2011). La plupart des RAB sont exprimées de manière ubiquitaire avec des niveaux d’expression différents selon les types cellulaires

(Martinez et Goud 1998). Ainsi RAB27a est préférentiellement exprimée dans les cellules de la lignée hématopoïétique, et RAB17 dans les cellules épithéliales. Cependant, certaines

RAB sont exclusives à des types cellulaires. Par exemple, RAB3 est uniquement exprimée dans les neurones où elle est associée aux vésicules synaptiques (Stettler et al. 1992; Fischer et al. 1990). Les RAB sont synthétisées dans le cytosol et sont associées à la face cytosolique des membranes de chaque compartiment cellulaire. Les domaines des RAB possédant le plus de variabilité sont les régions C-terminale et N-terminale. Dans le domaine C-terminal sont

44 45 présentes deux cystéines pouvant être prénylées (Hutagalung et Novick 2011). L’ajout de deux géranylgéranyl sur ces deux cystéines permet l’ancrage des RAB à la surface des membranes (Stenmark 2009a). Les RAB possèdent une structure tertiaire similaire à celle des autres GTPases avec six feuillets  entourés de 5 hélices  (Lee et Lambright 2009;

Pfeffer 2005). Elles présentent en leur centre une structure en boucle P qui permet de lier le nucléotide guanine et aussi, de capter un ion Mg2+ qui agit comme cofacteur dans la régulation des RAB (Sprang 1997). Deux régions, nommées switch I et II, proches du domaine de liaison au nucléotide guanine se voient subir des grands changements conformationnels lors de l’échange du nucléotide (Merithew et al. 2001). Ces domaines sont impliqués dans des interactions protéiques et régulent le recrutement d’effecteurs (Khan et

Ménétrey 2013).

1.3.1.3 Régulation des RAB

L’activité des RAB est régulée par l’échange du GDP par une molécule de GTP

(Armstrong 2000). En effet, sous sa forme liée au GTP, la RAB est considérée comme active.

Sous sa forme liée au GDP, la RAB est considérée comme inactive (Barr et Lambright 2010).

Lorsqu’elles sont associées à une membrane cible, les RAB sont activées par des GEF

(Guanine Nucleotide Exchange Factors) qui échangent le GDP pour un GDP (Schoebel et al. 2009). Lorsqu’elles sont liées au GTP, les domaines switch I et II changent de conformation permettant le recrutement d’effecteurs, médiant ainsi la fonction des RAB

(Pylypenko et al. 2018). Une fois leurs fonctions réalisées, les RAB sont inactivées par des

GAP (Ligeti et al. 2012), qui induisent l’hydrolyse du GTP en GDP. Ce changement de nucléotide induit de nouveau un changement de conformation des domaines switch I et II inhibant les interactions avec leur effecteur (Lee, et al. 2009). Ainsi, sous leur forme liée au

45 46

GDP, les RAB sont inactives et localisées dans le cytosol (Figure 8). Les RAB sont solubilisées grâce à l’interaction avec une GDI (GDP dissociation inhibitors) (Pylypenko et al. 2006). Par ailleurs, les GDI jouent un rôle dans le transport et l’ancrage des RAB à une membrane cible et ce, en réponse à un stimuli (Pfeffer and Aivazian 2004). Il a été montré, dans plusieurs cas, que les GEF peuvent aussi réguler l’arrimage des RAB à leur membrane cible (Blümer et al. 2013). Une fois ancrée à la membrane, les RAB sont activées par des

GEF (Lamber et al. 2019). Les RAB cyclent donc entre leur forme active et leur forme inactive et fonctionnent comme des interrupteurs moléculaires (Figure 8).

Figure 8 : Régulation cyclique des RAB

Les RAB sont prénylées en C-terminal permettant leur ancrage à la surface des membranes lipidiques. Les RAB liées au GDP sont inactives. Ce sont les GEF, en échangeant le GDP par un GTP qui permettent d’activer les RAB. Liées au GTP les RAB sont alors actives et recrutent des effecteurs afin d’accomplir leurs fonctions. Les RAB sont ensuite inhibées par une GAP qui active l’activité intrinsèque des RAB à hydrolyser le GTP en GDP. L’hydrolyse du GTP provoque un changement de conformation déstabilisant l’interaction avec les effecteurs. Sous leur forme inactive les RAB sont solubilisées par l’interaction avec une GDI qui par la suite joue un rôle dans l’adressage des RAB à une membrane spécifique. Les RAB subissent donc un cycle d’activation, d’inhibition, de solubilisation et d’adressage.

46 47

1.3.1.4 Les GEF

Les GEF interagissent avec les RAB et catalysent l’échange du GDP par un GTP (Bos et al. 2007). Au vu du grand nombre de RAB existantes, il existe plusieurs GEF afin d’assurer une spécificité. À ce jour, une trentaine de GEF ont été identifiées (Koch et al. 2016). Ces

GEF sont classées en différentes familles selon le type de domaine qu’elles possèdent. On peut noter les monomères VPS9 et DENN, l’homodimère SEC2, l’hétérodimère MON1-

CCZ1, et l’oligomère TRAPPI (Hutagalung et Novick 2011). Les mécanismes permettant l’échange du GDP ne sont pas conservés dans chacune de ces classes. Cependant et de manière générale, les GEF interagissent avec les domaines switch I et II (Itzen et al. 2006) des RAB compétitionnant de manière allostérique avec l’ion Mg2+ (Bourne et al. 1991).

L’interaction des GEF avec les RAB induit un changement de conformation de la RAB qui ouvre la poche de liaison du nucléotide (Pylypenko et al. 2018). L’excès de GTP cellulaire permet le déplacement du GDP. Le GTP se fixe dans la région de la boucle en P et induit l’activation des RAB par la stabilisation des domaines switch I et II (Vetter et Wittinghofer

2001). La spécificité d’une GEF pour une RAB est définie par sa capacité à interagir avec les régions switch I et II mais aussi par sa capacité à modifier leur conformation (Pylypenko et al. 2018). Malgré une certaine spécificité, une GEF peut activer plusieurs RAB et une RAB peut être activée par plusieurs GEF (Cherfils et Zeghouf 2013).

1.3.1.5 Les GAP

Les GAP interagissent avec les RAB liées au GTP et favorisent l’activité intrinsèque des RAB à hydrolyser le GTP en GDP (Pfeffer 2001). De la même manière que pour les

GEF, il existe de nombreuses GAP, 44 ont été identifiées et sont classées dans plusieurs

47 48 groupes. Aussi, une RAB peut être inhibée par plusieurs GAP et une GAP peut inhiber plusieurs RAB (Frasa et al. 2012). En revanche, l’ensemble des GAP identifiées, à ce jour, possèdent un domaine catalytique TBC (Tre-2/Cdc16/Bub2) (Pan et al. 2006). Les GAP interagissent aussi avec les domaines switch I et II des RAB. Elles sont composées de deux sous domaines qui entourent la région de liaison au GTP (Barr et Lambright 2010). Cela a pour effet de modifier la conformation de la RAB et d’activer son activité hydrolytique. Les

RAB peuvent être classées en deux classes selon leur mode d’action. Il y a les GAP dites classiques qui utilisent le mécanisme « dual trans-finger » ou RQ (Arginine-Glutamate), où une arginine conservée permet de stabiliser le phosphate du GTP (Pan et al. 2006). Chacune des combinaisons, RQ, RR, RX, QQ, et XQ sont retrouvées chez certaines GAP. Cependant, il existe des GAP ne possédant ni de motif d’arginine, ni de motif de glutamate (Pan et al.

2006). Dans ce cas, les GAP présentent généralement une asparagine avec une activité catalytique similaire.

1.3.1.6 Effecteurs et fonctions des RAB

Sous leur forme active, les RAB régulent chacune des étapes du trafic membranaire, à savoir le bourgeonnement d’une vésicule de la membrane du compartiment initial, le transport, l’ancrage et la fusion de cette vésicule avec la membrane du compartiment cible

(Zerial et McBride 2001). En effet, lorsque les RAB sont liées au GTP, cela a pour effet de stabiliser les structures des domaines switch I et II et donc de permettre l’interaction avec divers effecteurs (Pylypenko et al. 2018). Les effecteurs sont les relais qui permettent d’effectuer les fonctions des RAB (Tableau 1). Les effecteurs sont caractérisés par leur capacité à lier les RAB : ils possèdent un domaine RDB (RAB Binding Domain) (Khan et

Ménétrey 2013). Les effecteurs ont été identifiés via des approches génétiques, des systèmes

48 49 de double hybride et des approches de purification par affinité tel que les GST pull-down

(Gillingham et al. 2014).

RAB Effecteur GEF GAP Localisation

RAB1 OCRL1, p115 TRAPPI TBC1D20 RE-Golgi

RAB2 HOPS, RUND-1 N/A TBC1D1, TBC1D4,

TBC1D11, TBC1D20 RE-Golgi

RAB3 NOC2, RIM1, MADD RAB3GAP, TBC1D10D TGN

Rabphilin-3A

RAB4 RABPEP1 N/A TBC1D11 Endosomes précoces et de

recyclage

RAB5 RABEP1, OCRL1, RABGEF1, TBC1D2, TBC1D3, Endosomes précoces et GAPVD1 USP6NL VPS35, EEA1 vésicules de clathrine

RAB6 KIF20A N/A TBC1D11 Golgi

RAB7 HOPS, RILP, VPS35 MON1- TBC1D2A, TBC1D5, Endosomes tardifs CCZ1 TBC1D15 RAB8 RAB8IP, OCRL1 GRAB, TBC1D1 Golgi et TGN

RAB3IP

RAB9 TIP47 DENND2 N/A Endosomes tardifs

RAB10 KIF13A/B DENND4 TBC1D1, TBC1D4 Golgi et TGN

RAB11 RAB11FIP5, SH3BP5 N/A TGN et endosomes de recyclage ZFYVE27, PI4KIIIb

RAB14 OCRL1 DENND6 TBC1D1 Golgi et endosomes

précoces

RAB22 HOOK1 RABGEF1 N/A Golgi et endosomes

précoces

RAB27 MLPH, SLP4 MADD TBC1D10A, Mélanosomes

TBC1D10B

49 50

RAB Effecteur GEF GAP Localisation

RAB29 LRRK2, RAB8, N/A N/A TGN RAB11

RAB35 OCRL1, MICAL1 DENND1A N/A Endosome de recyclage

RAB39 MYO5A N/A N/A Golgi

Tableau 1 : Régulateurs, effecteurs et localisations des RAB Dans ce tableau sont représentés une liste non exhaustive des effecteurs, GEF et GAP associés à 17 RAB. Lorsque les données sont non disponibles une mention non applicable (N/A) est annotée. Aussi, les localisations majeures de chacune des RAB sont précisées. Il est important de noter que les localisations sont celles ou les RAB ont principalement été observées. En effet, chacune de ces RAB peuvent être associées à la membrane d’autres compartiments cellulaires selon le contexte. Cette liste a été établie à partir des données des deux articles de revue suivant : Bhuin et Roy 2014, et Muller et Goody 2018.

Une même RAB peut interagir avec plusieurs effecteurs et un même effecteur peut interagir avec plusieurs RAB. Par exemple OCRL1 (Tableau 1), une enzyme qui hydrolyse le phosphate de PIP, est l’effecteur de RAB1, RAB5, RAB6, RAB8, RAB14, et RAB35

(Hyvola et al. 2006). De la même manière, RAB5 peut recruter plusieurs effecteurs tels que

EEA1, RABEP1 (RAB GTPase Binding Effector Protein 1) et Rabenosyn-5 (RBSN)

(Stenmark 2009). Aussi, il a été montré que RAB11 peut interagir avec deux effecteurs,

Rabin8 et PI4KIII simultanément (Vetter et al. 2015). Enfin, des exemples montrent que certains effecteurs interagissent avec la forme inactive des RAB. En effet, la protrudine

(ZFYVE27) interagit avec RAB11 sous sa forme liée au GDP (Campa et Hirsch 2017). Les

RAB régulent chacune des grandes étapes du trafic membranaire, à savoir le bourgeonnement d’une vésicule, la fission, le transport et enfin l’ancrage et la fusion de cette vésicule avec la membrane d’un compartiment cible (Hutagalung et Novick 2011). Par exemple, RAB1, localisée au RE, régule la formation des vésicules COPII (Jedd et al. 1995). En effet, RAB1 recrute p115 qui est impliqué dans le recrutement de COPII et le bourgeonnement des

50 51 vésicules (Short et al. 2005). De plus, p115 recrute Rbet1 et Syntaxine 5, un complexe de

SNARE, qui permet l’ancrage et la fusion des vésicules COPII avec la membrane du Golgi

(Wang et al. 2015). D’autres données montrent que RAB5 est impliqué dans la formation des vésicules de clathrine issues de la membrane plasmique (Liu et al. 2017). Ensuite, les RAB peuvent interagir directement avec des protéines motrices qui permettent le transport de vésicules le long du cytosquelette. RAB6 interagit directement avec Rabkinesin-6 qui permet le transport de vésicules du Golgi au RE (Miserey-Lenkei et al. 2017). L’exemple le plus caractérisé est le cas de RAB27a qui interagit avec la myosine-Va qui permet le transport des mélanosomes le long du cytosquelette d’actine (Strom et al. 2002). L’ancrage de la vésicule avec la membrane du compartiment cible est un élément clé du trafic. En effet, avant d’induire la fusion des membranes, la vésicule doit être stabilisée à proximité du compartiment cible (Spang et al. 2016). Les protéines d’ancrage sont soit des protéines avec de longues répétitions de domaines en superhélice ou des complexes avec de grandes sous- unités (Eisenberg-Bord et al. 2016). Par exemple, RAB5 recrute directement EEA1 (Early

Endosome Antigen 1) pour médier l’ancrage et ensuite la fusion homotypique des endosomes précoces (Christoforidis et al. 1999). La fusion est régulée par les SNARE (Wang et al. 2017).

De la même manière, RAB5 recrute la SNARE, Syntaxin-13, et les effecteurs Rabaptin5 régulant ainsi la fusion des membranes des endosomes (Stenmark et al. 1995; McBride et al.

1999). Ainsi, une RAB peut recruter successivement plusieurs effecteurs impliqués dans le même processus et réguler tout un pan du trafic membranaire. Enfin, les RAB sont aussi impliquées dans la sélection des cargos. En effet, RAB7 recrute le CSC (VPS35-VPS26-

VPS29) du Rétromère aux endosomes (Priya et al. 2015). De plus, certaines RAB comme

RAB11 interagissent directement avec des cargos. Ainsi, RAB11 régule le transport antérograde du réseau transgolgien à la membrane plasmique de la protéine p14 FAST par

51 52 interaction directe (Parmar and Duncan 2016). Les RAB sont donc impliquées dans chacune des étapes de trafic membranaires de cargos.

1.3.1.7 Localisation des RAB

Les RAB, en condition basale, sont fortement associées avec la membrane d’un compartiment spécifique (Zerial et McBride 2001). Elles sont donc considérées comme des marqueurs membranaires. Ainsi, RAB1 et RAB2 sont localisées dans les compartiments entre le RE et Golgi (Figure 9). RAB1 y régule le trafic antérograde et RAB2 le trafic rétrograde (Tisdale et al. 1992). RAB6 et RAB8 sont présentes au réseau transgolgien et médient la sécrétion des protéines nouvelles synthétisées (Martinez et al. 1994; Peränen et al. 1996). RAB5 est impliquée dans la régulation des événements d’endocytose dépendant de la clathrine et des cavéolines (Liu et al. 2017; Zhang et al. 2018). RAB9 et RAB7 sont associées avec les endosomes tardifs et les lysosomes (Figure 9). RAB9 et RAB7 sont impliquées dans la maturation des endosomes tardifs et la fusion avec les lysosomes

(Vanlandingham et Ceresa 2009; Ganley et al. 2004; Luzio et al. 2004). RAB11 est présente aux endosomes de recyclage et favorise le recyclage à la membrane plasmique (Takahashi et al. 2012).

L’endosome précoce étant une gare de triage impliquée dans le triage des cargos endocytés vers l’ensemble des compartiments cellulaires, plusieurs RAB y sont donc présentes et y définissent des microdomaines (Jovic et al. 2010). En effet, RAB4, RAB5,

RAB22 et RAB21 sont localisées aux endosomes précoces (Wandinger-Ness et Zerial 2014).

RAB22 régule le transport rétrograde au réseau transgolgien (Kauppi et al. 2002), RAB4 le recyclage rapide à la membrane plasmique (H. Li et al. 2008) et RAB21 a été associée à plusieurs fonctions.

52 53

Figure 9 : Localisation des RAB Les RAB sont présentes à la membrane lipidique de chaque organite. RAB1 et RAB2 régulent les échanges entre le RE et le Golgi. RAB6 et RAB8 régulent le transport depuis le réseau transgolgien et permettent la formation de vésicules de sécrétion. RAB4, 5, 11 et 22 sont associées aux endosomes précoces. RAB4 régule le recyclage à la membrane plasmique, RAB11 le recyclage aux endosomes de recyclage, et 22 le transport rétrograde au Golgi. RAB5 est présent à la membrane plasmique où elle permet l’internalisation de cargos vers l’endosome précoce. RAB9 et RAB7 sont associées aux endosomes tardifs et médient la maturation des endosomes. RAB7 est aussi présente aux lysosomes.

1.3.2 RAB21

1.3.2.1 Généralités concernant RAB21

RAB21 est une petite GTPase de 24 kDa associée au trafic endosomal (Simpson et al.

2004). RAB21 fait partie de la famille des RAB et plus particulièrement de la sous famille associée à RAB5. En effet, RAB21 a pour plus proche homologue RAB5 (Pereira-Leal et

Seabra 2001). RAB21 est exprimée de manière ubiquitaire avec des niveaux variant d’un

53 54 type cellulaire à l’autre (Opdam et al. 2000). RAB21 est majoritairement associée au trafic endosomal mais des exemples montrent que RAB21 peut être localisée à la membrane plasmique, au Golgi et au réticulum (Burgo et al. 2009). Certains partenaires protéiques de

RAB21 dont des cargos et des effecteurs ont été caractérisés (Tableau 2). Par ailleurs, deux

GEF ont été identifiées pour activer RAB21, à savoir MTMR13 (Jean et al. 2012) et VARP

(Burgo et al. 2009). Cependant aucune GAP n’est connue pour inhiber RAB21.

Interacteurs Effecteurs GEF GAP Cargo Référence

APPL1 ? Zhu et al. 2007

EGFR X Simpson et al. 2004

ITGB1 X Pellinen et al. 2006

MTMR13 X Jean et al. 2012

PS1 X Sun et al. 2018

RABGEF1 ? Mori et Fukuda 2013

RPH3A X Yuan et al. 2017

TI-VAMP X Burgo et al. 2009

TRAF3-MKK3 X Li et al. 2019

TLR4 X Wang et al. 2019

VAMP8 X Jean et al. 2015

VARP X Burgo et al. 2009

Tableau 2 : Partenaires protéiques de RAB21 Dans ce tableau sont représentés l’ensemble des partenaires protéiques connus de RAB21. Une croix définit le type de relation entre RAB21 et chacun de ses partenaires. Lorsque le type de relation entre RAB21 et un de ces partenaires n’a pas été élucidé un point d’interaction indique le type d’interaction supposé.

54 55

1.3.2.2 Localisation de RAB21

Des études de microscopie confocale montrent que RAB21 est majoritairement localisée sur des structures associées au trafic endosomal (Simpson et al. 2004). En effet,

RAB21 colocalise avec des vésicules positives pour EEA1, un effecteur de RAB5 et marqueur des endosomes précoces (Simpson et al. 2004). De plus, des études comparatives de localisation montrent que RAB21 est associée aux mêmes structures que RAB4, RAB5 et

RAB22, des RAB associées aux endosomes précoces (Simpson et al. 2004). Enfin, des analyses de microscopie à fluorescence montrent que le traitement de cellules avec la wortmannin, un inhibiteur de PI3K, induit la formation de tubules émergeant des endosomes positifs pour RAB5 (Egami et Araki 2008). Ces tubules sont très dynamiques, dépendent des microtubules et sont positives pour RAB21. La localisation endogène de RAB21 aux endosomes précoces a été confirmée dans des lignées d’hépatome (Simpson et al. 2004). En revanche RAB21 colocalise faiblement avec les RAB associées aux endosomes tardifs et lysosomes, RAB7 et RAB9. Cette localisation est aussi observée dans le parasite

Trypanosome Brucei où RAB21 colocalise partiellement avec TbVps23, un marqueur des endosomes tardifs (Ali et Field 2014). De plus l’inhibition de RAB21, dans le trypanosome, inhibe le triage vers les lysosomes (Ali et Field 2014). Au contraire de la forme sauvage de

RAB21, la forme dominante négative de RAB21, c’est-à-dire liée au GDP induite par la mutation T33N, est localisée au RE et au Golgi (Simpson et al. 2004). De manière intéressante il a été montré que RAB21 est localisée au niveau de structures réticulaires dans des cellules Caco2 non polarisées et au niveau de vésicules apicales dans les cellules Caco2 polarisées (Opdam et al. 2000).

55 56

Ainsi, la localisation de RAB21 semble être affectée par son état d’activation, le type cellulaire, et l’état de différentiation cellulaire. Ces données supposent que plusieurs fonctions, dépendant de signaux cellulaires, sont associées à RAB21. En effet, des données suggèrent que sous certaines conditions RAB21 régule des évènements d’endocytose à la membrane plasmique, autre localisation potentielle de RAB21 (Figure 10).

1.3.2.3 Fonctions de RAB21

De par les études décrites ci-haut, RAB21 est principalement localisée aux endosomes précoces et les premières études montrent que RAB21 est impliquée dans la fusion de ces endosomes (Zhang et al. 2006). En effet, la surexpression du dominant négatif de RAB21

(T33N), inhibe la capacité de fusion des endosomes précoces. Ensuite, plusieurs exemples de la littérature attribuent un effet de RAB21 dans la régulation de l’endocytose de plusieurs cargos.

Tout d’abord, il a été montré que RAB21 régule l’internalisation des intégrines 1

(Figure 10). En effet, RAB21 interagit directement avec les sous-unités 1, 2 et 5 des hétérodimères d’intégrine liés à la sous-unité 1 et régule l’internalisation des intégrines présentes aux points focaux d’adhésion (Pellinen et al. 2006). Ainsi, RAB21 régule l’adhésion et la migration cellulaires. En effet, deux agents pharmacologiques le Lovastatin et le Simvastatin perturbent le remodelage de la matrice extracellulaire (MEC) et réduit la capacité d’invasion des cellules de carcinome squameux (Squamous Carcinoma Cells). Ces composés perturbent l’activité de RAB dont celle de RAB21 (Hooper et al. 2009).

56 57

Figure 10 : Principales fonctions associées à RAB21 RAB21 est principalement localisée aux endosomes précoces. L’activation de RAB21 par MTMR13 permet d’induire le triage de VAMP8 vers les lysosomes ce qui a pour effet de médier la fusion entre autophagosomes et lysosomes, étape clé de l’autophagie. Aussi, en réponse à une stimulation à l’EGF, RAB21 induit l’internalisation de l’EGFR et son triage vers la dégradation lysosomale. Ici RAB21 a un effet d’inhibition de l’activité de la voie EGF. Enfin, RAB21 régule l’internalisation des intégrines ce qui a pour effet d’induire l’activation de la voie FAK au niveau des vésicules d’endocytose. RAB21 est impliquée dans l’internalisation et le triage vers la dégradation d’autres cargos qui ne sont pas représentés ici par soucis de clarté.

57 58

Dans ce contexte RAB21 est donc requise pour l’accumulation des intégrines 5 à la membrane plasmique permettant le remodelage de la MEC. RAB21 ne régule pas seulement l’endocytose des intégrines. Une étude montre que RAB21 régule le triage des intégrines au fuseau mitotique lors de la cytokinèse (Pellinen et al. 2008). L’absence de RAB21 est associée avec des défauts de cytokinèse et induit l’apparition de cellules polynucléées. Enfin, il a été montré que l’internalisation des intégrines 1 par RAB21 permet l’activation de la voie FAK (Focal Adhesion Kinase 1) au niveau des endosomes, ce qui a pour effet d’inhiber l’anoïkose, mort par détachement cellulaire (Alanko et al. 2015). En effet, l’inhibition de l’expression de RAB21 inhibe la phosphorylation de FAK et induit la mort des cellules lorsque celles-ci sont mises en suspension (Alanko et al. 2015). Plusieurs exemples montrent que RAB21 interagit avec l’EGFR (Simpson et al. 2004). La surexpression de RAB21 entraine l’internalisation et le triage vers la dégradation de l’EGFR et ce, en présence ou en absence de stimulation à l’EGF (Figure 10). De plus, la surexpression de RAB21 inhibe la voie des MAPK (Mitogen-Activated Protein Kinase 1) induite par l’EGF (Yang et al. 2012).

Les études récentes suggèrent que RAB21 est impliquée dans la régulation de la réponse inflammatoire (Li et al. 2019). En effet, l’inhibition de RAB21 dans des BMDM (Bone-

Marrow-Derived Macrophage), des cellules de macrophage de souris, inhibe la synthèse de cytokines pro-inflammatoires telles qu’IL-6 (Interleukine), IL-1, et TNF (Tumor Necrosis

Factor) et ce, suivant une stimulation au LPS (Lipopolysaccharides), un agent inflammatoire.

De même, la perte d’expression de RAB21 inhibe l’internalisation du TLR4 (Toll-Like

Receptor 4) et ainsi, diminue l’activation des voies c-Jun et NF-B (Nuclear Factor-kappa

B) (Li et al. 2019). Enfin, la surexpression de RAB21 induit l’endocytose et le triage de la préséniline 1 (PS1) vers les lysosomes. PS1 est une des sous-unités catalytiques de la -

58 59 sécrétase impliquée dans la synthèse de l’amyloïde  (Sun et al. 2018). RAB21 a pour effet de favoriser la synthèse de l’amyloïde  alors que l’inhibition de RAB21 réduit l’accumulation de l’amyloïde  (Sun et al. 2018). RAB21 est donc présente à la membrane plasmique où elle régule l’endocytose de plusieurs cargos ayant un effet sur leurs niveaux d’activité. Cependant, des données montrent que l’inhibition de l’expression de RAB21 ne semble pas affecter les niveaux d’activation de la voie EGF supposant que les effets associés

à RAB21 sont dépendants du type cellulaire et du contextes (Jean et al. 2015).

Les exemples cités ci-dessus, pour PS1 et l’EGFR, suggèrent un rôle pour RAB21 dans des évènements de triage de l’endosome précoce vers le lysosome. En condition de carence nutritionnelle, RAB21 est activé par la GEF MTMR13 et induit le triage de VAMP8 des endosomes précoces vers les lysosomes (Jean et al. 2015). VAMP8 est une SNARE qui associée avec SNAP29 et STX17 régule la fusion du lysosome avec l’autophagosome (Liu et al. 2015). Cette étape de fusion est une étape limitante dans la régulation du flux autophagique. En effet, ce n’est que lors de cette fusion que les composants cytosoliques séquestrés au sein de l’autophagosome peuvent être dégradés par les enzymes lysosomales

(Saha et al. 2018). RAB21 joue donc un rôle clé dans la régulation de l’autophagie induite par la carence et l’inhibition de l’expression de RAB21 induit l’accumulation d’autophagosome en condition de carence. De manière intéressante, l’expression de la forme active de RAB21 favorise aussi les évènements de phagocytose (Khurana et al. 2005). Si l’autophagie est le mécanisme de dégradation de composants intracellulaires par les lysosomes, la phagocytose est son pendant pour des composants extracellulaires. La localisation de RAB21 au Golgi est associée à sa forme liée au GDP. Cependant, il a été montré que VARP, TI-VAMP et RAB21 sont localisées dans des régions périnucléaires. En

59 60 effet, VARP est une GEF qui active RAB21 qui régule le triage de TI-VAMP et joue un rôle dans la formation de neurites lors de la différentiation de l’hippocampe (Burgo et al. 2009).

Ces données supposent que RAB21 peut être activée au Golgi et avoir une activité impliquée dans la régulation de ces évènements cellulaires.

1.3.2.4 Rôle de RAB21 dans des pathologies

Au vu du rôle de RAB21 dans la régulation des voies de signalisation associées à la survie, la prolifération et la migration, il est possible que RAB21 soit impliquée dans des processus tumorigènes. Ainsi, l’inhibition de RAB21 inhibe la croissance cellulaire et induit la mort apoptotique des lignées cellulaires cancéreuses T98G et U87 (Ge et al. 2017).

L’inhibition de RAB21 induit l’expression de protéines pro-apoptotiques (caspase 7, Bim et

Bax) dans ces cellules de gliome. De plus, il a été montré que l’inhibition préalable de l’expression de RAB21 dans des cellules MDA-MB-231, issues d’un adénocarcinome mammaire, injectées dans des souris a pour effet d’inhiber la formation de métastases

(Alanko et al. 2015). Cependant, l’inhibition de RAB21 dans des lignées cancéreuses colorectales semble avoir des effets variables sur leur prolifération (Lauzier et al. 2019).

RAB21 semble donc avoir des effets ambigus dans des processus carcinogènes qui semblent dépendants du contexte cellulaire. Considérant les fonctions de RAB21 dans l’internalisation de PS1, de TLR4, et dans la régulation de l’autophagie, il est possible que RAB21 soit impliquée dans le développement de maladies neurodégénératives et dans la réponse inflammatoire, respectivement.

60 61

Hypothèse

Le trafic membranaire régule l’ensemble des échanges entre les différents compartiments cellulaires ainsi qu’entre les domaines intracellulaire et extracellulaire. Ce trafic est essentiel à l’homéostasie cellulaire et hautement régulé. Au sein de ce trafic, l’endosome précoce joue un rôle central dans le triage des cargos endocytés vers les différents compartiments cellulaires. Le trafic de récepteurs soit vers la dégradation soit vers le recyclage a des effets importants sur les niveaux d’activité de voies de signalisation. Il existe donc de nombreux complexes protéiques présents aux endosomes précoces, tels que le complexe WASH et le Rétromère qui régulent ces évènements de triage. L’activité de ces complexes est régulée par les RAB qui agissent comme les chefs d’orchestre du trafic membranaire. Plusieurs RAB sont localisées aux endosomes précoces dont RAB21. Cette dernière est une petite GTPases énigmatique impliquée dans la régulation du triage endosomal mais aussi dans des évènements d’endocytose et dans la régulation de mécanismes cellulaires liés au Golgi. De plus, RAB21 joue un rôle essentiel dans la régulation de mécanismes impliqués dans la prolifération, la migration et la survie cellulaire.

RAB21 doit donc être régulée de manière différentielle selon le contexte cellulaire.

Cependant seules trois GEF ont été identifiées pour RAB21, et à ce jour, aucune GAP et peu d’effecteurs n’ont été mis en évidence pour inhiber et médier l’activité de RAB21.

Nous émettons donc l’hypothèse que les fonctions spécifiques associées à RAB21 sont définies et régulées par la complexité de son interactome. Le projet consiste donc à identifier l’ensemble des partenaires protéiques de RAB21 que les approches classiques d’études des partenaires protéiques des RAB n’ont pas permis de caractériser.

61 62

Objectifs

Objectif #1 Identifier les partenaires de RAB21 par la combinaison des approches APEX2 et de purification par affinité (AP) combinée à la spectrométrie de masse (MS), et valider les interactions identifiées. Objectif #2 Valider l’interaction identifiée et caractériser la relation fonctionnelle entre RAB21 et TMED10

62 63

2ARTICLE 1 APEX2-mediated RAB proximity labeling identifies a role for RAB21 in clathrin- independent cargo sorting

Auteurs de l’article: Del Olmo T, Lauzier A, Normandin C, Larcher R, Lecours M, Jean D, Lessard L, Steinberg F, Boisvert FM, Jean S.

Statut de l’article: publié dans EMBO Report 2019 Feb;20(2). pii: e47192. doi: 10.15252

Avant-propos: Pour ce projet, j’ai généré les lignées APEX2:RAB21 et GFP:RAB21WT/CA/DN. J’ai validé l’expression et l’activité d’APEX2:RAB21 et les localisations des différentes formes de GFP:RAB21 ainsi que de l’induction de biotinylation par des expériences d’immunofluorescence et de microscopie confocale. J’ai ensuite réalisé l’ensemble des expériences de spectrométrie de masse ainsi que leur analyse. Par la suite, j’ai validé l’interaction entre RAB21 et les membres des complexes WASH et Rétromère par des expériences de co-immunoprécipition et de GST-pull down. Enfin, j’ai mesuré les différents niveaux d’expression de RAB5, RAB7, VPS35, VPS29, VPS26, Strumpellin, WASH, FAM21 et CAPZ par des expériences d’immunobuvardage. Concernant la rédaction de l’article, j’ai dessiné plusieurs schémas explicatifs et j’ai participé à la mise en page de figures. Aussi j’ai participé à l’écriture de certaines parties du texte. Ma contribution à cet article représente donc environ 65% du travail présenté.

Résumé : Les RAB GTPases sont des modulateurs centraux du trafic membranaire. Elles sont activées par des GEF (Guanine Exchange Factor) et inhibées par des GAP (GTPase Activating Protein) de manière cyclique. Une fois activée, les RAB recrutent un large spectre d'effecteurs pour contrôler les fonctions de trafic des cellules eucaryotes. Les approches classiques de pull-down ou de double-hybride ont identifié un certain nombre d'interacteurs des RAB. Toutefois, en raison de la nature de ces approches et des limites inhérentes à chaque technique, l’identification des partenaires des RAB demeure incomplète. En comparant les enrichissements protéiques obtenus via le marquage de proximité par APEX2 des RAB4a, RAB5a, RAB7a et ceux obtenus avec GFP:RAB21, nous avons constaté que l'approche APEX2 est très efficace pour l'identification globale des effecteurs et régulateurs des RAB. Il est important de noter que, grâce à des approches biochimiques et génétiques, nous avons pu établir un nouveau lien entre RAB21 et les complexes WASH et Rétromère, avec des conséquences fonctionnelles sur le tri de cargos. Par conséquent, le marquage par proximité des protéines voisines des RAB par APEX2 représente un nouvel outil très efficace pour définir les fonctions des RAB.

63

64

APEX2-mediated RAB proximity labeling identifies a role for

RAB21 in clathrin-independent cargo sorting

Tomas Del Olmo1, Annie Lauzier1, Caroline Normandin1, Raphaëlle Larcher1, Mia Lecours1,

Dominique Jean1, Louis Lessard1, Florian Steinberg2, François-Michel Boisvert1 and Steve

Jean1*

*Corresponding author:

Email: [email protected]

Telephone: 819-821-8000 Ext: 70450

FAX: 819-820-6831

1Faculté de Médecine et des Sciences de la Santé

Department of Anatomy and Cell Biology

Université de Sherbrooke

3201, Rue Jean Mignault

Sherbrooke, Québec, Canada, J1E 4K8

2Center for Biological Systems Analysis (CBSA)

Faculty of Biology, Albert Ludwigs Universitaet Freiburg

Freiburg, Germany

Running title: RAB21 regulates CIE cargo sorting

64 65

Keywords: RAB GTPases, APEX2, WASH complex, Retromer, Clathrin-Independent

Endocytosis

Abstract

RAB GTPases are central modulators of membrane trafficking. They are under the dynamic regulation of activating guanine exchange factors (GEF) and inactivating GTPase activating proteins (GAP). Once activated, RABs recruit a large spectrum of effectors to control trafficking functions of eukaryotic cells. Multiple proteomic studies, using pull-down or yeast two-hybrid approaches, have identified a number of RAB interactors. However, due to the in vitro nature of these and to inherent limitations of each technique, a comprehensive definition of RAB interactors is still lacking. By comparing quantitative affinity purifications of GFP:RAB21 with APEX2-mediated proximity labeling of RAB4a, RAB5a, RAB7a and

RAB21, we find that APEX2 proximity labeling allows for the comprehensive identification of RAB regulators and interactors. Importantly, through biochemical and genetic approaches, we establish a novel link between RAB21 and the WASH and retromer complexes, with functional consequences on cargo sorting. Hence, APEX2-mediated proximity labeling of

RAB neighboring proteins represents a new and efficient tool to define RAB functions.

65 66

Introduction

Membrane trafficking, the vesicular transport of cellular constituents, is fundamental to cellular and organismal homeostasis [1-3]. The sequential transport of cargos between intracellular compartments occurs through multi-step processes that are actively regulated

[4]. One important class of regulators controlling vesicular traffic and membrane compartment identity is the RAB GTPase family [5].

RAB GTPases form the largest family of small GTPases with nearly 70 members in humans

[6]. RABs are involved at every step of vesicular trafficking where they regulate sorting, fission, transport, tethering and fusion events [7]. RABs cycle between their inactive (GDP bound) and active (GTP bound) states through the action of Guanine Exchange Factors

(GEFs), and are conversely converted from their GTP bound state to their GDP state by

GTPase Activating Proteins (GAPs) [8]. RABs are often targeted by multiple GEFs and

GAPs [9,10]. The differential actions of these GEFs or GAPs can potentially direct a RAB to alternate compartments or specify particular functions [11]. Once recruited and activated at their target membranes, RABs bind effectors in order to mediate various functions [12].

RAB effectors recruit multiple classes of proteins ranging from adaptor coat proteins, to myosins, dyneins and signaling kinases [12]. Moreover, RABs also interact with various cargos, tethering complexes and cargo sorting complexes [13]. One such example is the role of RAB7 in the recruitment of the retromer complex at endosomes [13,14]. The retromer complex is an evolutionary conserved complex [15] which regulates endosome-to-Golgi retrograde transport [16] as well as endosome-to-plasma membrane trafficking [17]. The core retromer complex is constituted of two subcomplexes, namely the VPS and SNX complexes.

The VPS complex (or cargo sorting complex) comprises VPS26, VPS29 and VPS35 [18],

66 67 and interacts with various cargos either directly or through binding with various SNXs and accessory proteins [17,19,20]. The SNX subcomplex interacts with endosomes through

PtdIns(3)P [21] and also modulates cargo sorting [22,23] as well as membrane tubulation

[24]. In addition, the retromer interacts with the WASH complex, through an interaction between VPS35 and the WASH complex subunit FAM21 [19,25,26]. The WASH complex generates F-actin at endosomes to mediate cargo sorting and tubule scission [27-30]. It is now acknowledged that the retromer/WASH complex forms heterogeneous ‘subcomplexes’ with various interacting proteins to specifically and temporally regulate a large array of cargos [17,20,31-33]. Given the role of RAB GTPases in membrane traffic, it is likely that other RABs are involved in various aspects of retromer/WASH functions.

RAB interactors have mostly been defined through in vitro pull-down and yeast two-hybrid approaches [34-37]. Pulldown experiments are a powerful tool to assess direct binding between a RAB and a specific protein, and have been successfully used to identify interactors for multiple RABs [38]. However, they preclude identification of context-specific RAB

GEFs, GAPs or effectors, since the temporal aspects of effector recruitment are lost. On the other hand, yeast-two hybrid allows for the identification of RAB GEFs, GAPs and effectors.

However, yeast two-hybrid only monitors binary interactions and, as a result, complexes interacting with RABs through multiple proteins cannot be identified. Recently, a pull-down based RAB-interactome screen was performed in Drosophila enabling identification of a large number of RAB effectors [37]. Unfortunately, harder to purify RABs showed very limited number of interactors [37]. Ongoing proteome-wide studies aimed at defining the human proteome [39] have also tested numerous RAB GTPases. Unfortunately, these studies used C-terminal tags for affinity purifications which are not appropriate for RABs, due to

67 68

RAB C-terminal prenylation. The suboptimal tagging of the RABs in these studies yielded a low number of interactors for most of the twenty-five RABs tested. This was particularly evident given the strong prevalence for enzymes linked to RAB prenylation in the interactome (i.e. CHM, CHML, RABGGTA/B) [39]. Hence, in order to understand how

RABs exhibit different cellular functions, it is imperative to accurately and extensively define their associated proteome in the appropriate setting. In an effort to develop new approaches to map RAB GTPase interactors, we have combined quantitative mass spectrometry and

APEX2 proximity labeling techniques. Herein, we describe APEX2-mediated proximity labeling as a new highly efficient method to rapidly map RAB regulators/effectors. This approach notably allowed defining a novel RAB21 interaction with the WASH/retromer complexes and a RAB21 role in endosomal sorting of a subset of clathrin-independent cargos.

68 69

Results

Quantitative mass spectrometry defines potential RAB interactors

Early endosomal RAB21 has well-described roles in mediating integrin internalization to control cell migration, anoïkis resistance and cell division [40,41]. RAB21 also regulates aspects of VAMP7 and VAMP8 trafficking to control neurite growth and autophagy respectively [42,43]. Unfortunately, a low number of specific interactors were identified for

RAB21 in a recent pull-down study [37]. Given the importance of RAB21-associated functions, and the difficulty in identifying RAB21 interactors through conventional approaches, RAB21 represents a good RAB on which to establish novel methodologies aimed at identifying RAB-associated proteins. Hence, a quantitative SILAC-based affinity purification (AP-MS) approach was devised to map RAB21 interactors.

The Flp-In/T-REx system [44] was used to generate stable HeLa and HCT116 cell lines expressing N-terminally GFP-tagged wild type (WT), GTP-locked (Q78L) or GDP-locked

(T33N) forms of human RAB21. Various RAB21 variants were chosen in order to maximize the recovery of GEFs, GAPs and effectors. GFP:RAB21 variants were expressed and properly localized in both HeLa and HCT116 cells, except for T33N which showed weaker early endosomal localization with a concomitant Golgi relocalization (Fig. EV1A to D), consistent with earlier findings [45].

Duplicate SILAC experiments for each GFP:RAB21 variants identified a vast spectrum of potential direct and indirect RAB21 interactors (Fig. EV1E, Fig. EV2A to C, Dataset EV1).

Surprisingly, functional annotations in Reactome of individual cell lines did not show an enrichment toward membrane trafficking (Fig. EV2D and E), and were reminiscent of those

69 70 observed from a recently-published GST:RAB21 interactome pull-down study (Fig.

EV2F)[46]. Since the interactome was generated in two cell lines, we hypothesized that proteins concomitantly enriched from both cell lines would likely represent RAB21 interactors. Hence, the HeLa and HCT116 RAB21 networks generated with all of the various

RAB21 variants were merged. This resulted in 29 proteins present altogether (Fig. EV1E), which interacted to different degrees with the WT, Q78L or T33N RAB21 variants (Fig.

EV2C, Dataset EV1). When organized into a network (Fig. EV1F), multiple membrane trafficking regulators or potential cargos were observed. Reactome pathway analysis revealed an enrichment in membrane and vesicular transport processes (Fig. EV1G), strengthening these core proteins as prospective RAB21 interactors. While this quantitative

AP-MS interactome yielded a highly relevant network of RAB21 binding proteins, it was nonetheless variable between repeats and cell lines and required the use of GDP/GTP locked constructs. Therefore, other approaches were sought to identify RAB regulators and interactors.

APEX2:RABs are properly localized in HeLa cells

One caveat of AP-MS and pull-down approaches is that cell lysis influences protein-protein interactions. To circumvent this issue, APEX2:RAB fusions were used to perform proximity labeling (Fig. 1A) [47]. APEX2 is an engineered peroxidase that biotinylates proteins, mostly on tyrosine [48], in a 10-20 nm radius [49]. It uses biotin-phenol as its substrate and the reaction is catalyzed by a one-minute H2O2 treatment, enabling the rapid covalent labeling of neighboring proteins.

70 71

A concern with proximity labeling approaches is that they often lead to the identification of large arrays of proteins [50]. We rationalized that by using three early endosomal RABs as baits we could 1) validate the technique, 2) identify general and specific early endosomal neighbors for the three RABs and 3) refine the RAB21 proteome established by AP-MS.

Moreover, given the known segregation of RAB5 and RAB4 at endosomes [51,52], we hypothesized that APEX2-labeling would also identify RAB-specific microdomains (Fig.

1B). APEX2, APEX2:RAB21, APEX2:RAB5a and APEX2:RAB4a HeLa Flp-In/T-REx cells were thus generated. APEX2 fused RABs were able to biotinylate endogenous proteins.

Strong biotinylation was observed in the presence of H2O2 and biotin-phenol, whereas weak biotinylation was noted when H2O2 was omitted (Fig. 1C). Immunofluorescence analyses revealed that biotinylated proteins in APEX2:RAB21, RAB5 and RAB4 cells localized on

EEA1 positive puncta and also in a cytosolic/reticular pattern, while APEX2-only cells showed diffuse cytosolic and nuclear staining with no EEA1 colocalization (Fig. 1D).

Together, these results indicate that APEX2 is active when fused with RABs and that

APEX2:RAB can biotinylate proteins at EEA1 positive endosomes.

APEX2:RAB proximity labeling is enriched for trafficking regulators/effectors

Three label-free independent proximity labeling experiments for each RABs and the APEX2 control were performed. In order to statistically identify interactors/neighbors and to remove cytosolic/reticular contaminants, the SAINT software was used for filtering and the ProHits- viz suite for data representation [53,54]. Using a very strict SAINT filtering score of 0.95, which corresponds to a false discovery rate of  1%, 1173 proteins were identified in

APEX2:RAB21, 819 in APEX2:RAB5 and 469 in APEX2:RAB4 when normalized to

71 72

APEX2 only (Fig. 2A and Dataset EV2). Importantly, biotinylation was equivalent between

APEX2:RABs (Fig. EV3A). Therefore, the different number of interactors/neighbors between the various RABs most probably reflects different degrees of proximity with endosomal proteins, rather than variable protein expression levels between baits. RAB5 and

21 showed 57% overlap between their neighboring proteins (666 proteins), while

RAB21/RAB4 and RAB5/RAB4 had 35% (412 proteins) and 51% overlap (415 proteins) respectively (Fig. 2A). When all RAB21, RAB5 and RAB4 neighbors were analyzed for (GO) term enrichment, molecular functions related to trafficking events were highly significant and overlapped extensively (Fig. 2B). Of significance, the APEX2:RAB approach highlighted cellular functions related to trafficking, something that both GST pull-down or

AP-MS approaches failed to achieve without heavy filtering and the use of numerous RAB variants and cell lines.

Wild type APEX2:RABs identify known GEFs, GAPs and effectors

Given the large number of proteins identified in the APEX2:RAB21, 5 and 4 datasets, the type of neighbors detected through this methodology was explored. A comparative analysis of the various RABs (Fig. 2C) on ProHits-viz was performed. Strikingly, GEFs, GAPs and effectors/interactors were identified from wild type RABs with mostly specific enrichments toward their predicted target. Using published interactors (without exhaustively listing them all, Fig. 2C, Dataset EV2), preferential associations between RAB21 and MTMR2 were observed, as shown in Drosophila [55], and with VAMP7 [42]. An association between

RAB21 and integrin(s) was also detected, although integrins were rather variable in their proximity with early endosomal RABs, as could be expected from previous reports [56].

APPL1/2 and EEA1, which are known RAB21 and RAB5 effectors, were also highly

72 73 recovered with both RAB21 and RAB5, while weak in RAB4. APEX2:RAB5-mediated proximity ligation also identified other known RAB5 effectors, including VPS34 and

KIF13A. It also enriched HOOK1/3, UHRF1BP1L, CC2D1A, FAM160A1 and Ccdc128, which represent previously identified RAB5 effectors [37]. Surprisingly, Rabenosyn-5, which binds both RAB5 and 4, was mostly enriched with RAB5 compared to RAB4.

RABIP4, RAB11FIP11 and VPS45, which are functionally mostly linked to RAB4, were also more prominently found with RAB5 or RAB21 compared to RAB4, while only GRASP-

1 was notably observed with RAB4. Nonetheless, all of these known RAB4 interactors, although less abundant with RAB4, were recovered at an <1% FDR in APEX2:RAB4 experiments.

The APEX2 approach also identified known RAB21 and RAB5 GEFs (Fig. 2C).

VARP, a RAB21 GEF, was highly abundant with RAB21, while the more general RABGEF1 was observed at a low FDR (<1%) with RAB21 and 5. RAB GAPs were also recovered by

APEX2:RABs (Fig. 2C). Of note, TBC1D2, previously described with RAB5 GAP activity in C. elegans [57], was abundant with RAB5, while GAPCENA, a known RAB4 GAP, was present with RAB4 [58]. No RAB21 GAPs have been described in the literature, although

TBC1D17 was shown to have weak catalytic activity toward RAB21 in vitro [58]. TBC1D4 and/or TBC1D15, a close relative of TBC1D17, could potentially act on RAB21 given their respective high abundance with RAB21.

To confirm that APEX2:RAB-identified proteins were recovered due to their close proximity to RABs, and not merely the result of their general endosomal localization, the

RAB21 dataset was compared to a previously published APEX2:2xFYVE dataset [59]. The

73 74 data was normalized using ProHits-viz and compared. After analysis, most proteins were only found with RAB21 (Fig. EV3B), while those shared between the 2xFYVE probe and

RAB21 were, for the most part, known PtdIns(3)P binding or associated proteins (Fig.

EV3C). Thus, proteins identified through APEX2:RAB were not simply observed due to their endosomal localization. To further ensure the specificity of the APEX2:RAB-identified proteins, early endosomal RABs were compared to APEX2:RAB7a, a late endosomal RAB.

APEX2:RAB7 localized appropriately and led to biotinylation at late endosomes/lysosomes as detected with LAMP1 (Fig. EV3E). APEX2:RAB7-identified proteins had a limited overlap with early endosomal RABs as expected (Fig. 2A to C and Fig. EV3D). Importantly, the well described RAB7 GAP TBC1D5 [14] and the known RAB7 interactor VPS35 [60] were highly enriched with RAB7 (Fig. 2C). Altogether, these experiments strengthen

APEX2:RAB proximity labeling as an efficient method to identify RAB interactors/neighbors.

Another important aspect of the APEX2:RAB proximity labeling approach was that it successfully enriched whole protein complexes with single RABs (Fig. 2C). This approach, using a single purification step, effectively identified all subunits of the EARP complex with

RAB4. This complex was recently identified as an early endosomal sorting complex interacting with RAB4 [61,62]. EARP is highly similar to the Golgi GARP complex which does not interact with RAB4 [37]. Of note, all EARP-specific subunits were identified with strong enrichment with RAB4, while the GARP complex-specific subunit was not recovered.

Also, the WASH [63] complex was strongly enriched with RAB21, while a few subunits were observed with RAB5 and RAB7 (Fig. 2C). Interestingly, VPS retromer subunits [18] were strongly enriched with RAB7, compared to early endosomal RABs (Fig. 2C), as

74 75 expected from previous work [13,14]. Although, early endosomal RABs were statistically enriched with the VPS retromer subunit, they showed stronger proximities to various SNXs and other retromer subunits (i.e. SNX2, SNX6).

RAB GTPases are known to be functionally linked to phosphoinositides [5]. Thus, their proximity to phosphoinositide regulatory enzymes was investigated. RAB21 and RAB5 were found to interact or be in close proximity to a wide range of phosphoinositide phosphatases and kinases (Fig. EV4A). In accordance with previous studies [64], APEX2- proximity labeling identified INPP4a and OCRL as strong RAB5 interactors/neighbors.

Interestingly, multiple Myotubularin family members were specifically identified with

RAB21. Moreover, PIKFYVE was also exclusively found in APEX2:RAB21 cells (Fig.

EV4A). These associations imply that RAB21 could potentially impact three phosphoinositide pools, namely PtdIns(3)P, PtdIns(5)P and PtdIns(3,5)P2. Altogether,

APEX2:RAB proximity labeling allowed identifying a wide array of known regulators/interactors of RAB GTPases.

APEX2:RABs identify novel RAB interacting proteins.

Most of the aforementioned proteins represent known interactions and complexes.

Consequently, the extent of the interactors/neighbors identified by the APEX2 technique validated the APEX2 approach. However, in addition, we also assessed whether

APEX2:RABs proximity labeling could identify novel regulatory interactions. In order to identify RAB21 interactors/neighbors, APEX2:RAB21 enriched baits were filtered (Fig.

EV4B). This led to the identification of i) PLEKHM2, which has known roles in lysosomal positioning and autophagy [65], ii) SLC7A11, which interacts with SLC3A2 to control amino

75 76 acid transport [66] and iii) USP7, a recently-identified WASH complex modulator [67].

Other identified RAB21 interactors/neighbors showed potential roles in TGF-B, adherent junctions and in cohesin functions. To assess whether proteins detected by mass spectrometry could be validated by other approaches, APEX2 and APEX2:RAB21 proximity labeling in

GFP:SARA transfected cells was performed. After cell lysis and GFP:SARA immunoprecipitation, biotinylated SARA was only observed in APEX2:RAB21 cells (Fig.

EV4C). Moreover, co-immunoprecipitation (coIP) experiments performed between

FLAG:RAB21WT and HA:PLEKHM2 or HA:SLC7A11 expressing cells allowed highlighting reproducible interactions between RAB21 and PLEKHM2 or SLC7A11 (Fig.

EV4D). These examples further strengthen the APEX2-proximity labeling technique and corroborate these newly established neighbors as novel RAB21 interactors.

RAB21 interacts and colocalizes with the WASH and retromer complexes

In order to further build from the APEX2 proximity labeling approach and define novel molecular functions for RAB21, its proximity to the WASH and retromer complexes was studied in further detail. Mass spectrometry data was confirmed by performing an anti-biotin immunoprecipitation [68] on APEX2:RAB expressing cells and the presence of WASH- and retromer-associated proteins tested. Using this approach, Strumpellin was more prevalent with APEX2:RAB21, while similar amounts of VPS35 between RAB5 and RAB21 (Fig. 3A) were detected. This trend is in accordance with the mass spectrometry approach (Fig. 2C).

Specific interactions between RAB21 and WASH/retromer complexes subunits were further tested by coIP. Endogenous VPS26, FAM21, Strumpellin and VPS35 immunoprecipitated with GFP:RAB21 (Fig. 3B and C). Although weak, these interactions were highly

76 77 reproducible and associated to different degrees with the various RAB21 variants. This is similar to RAB21 association with integrins [69]. Of particular note, VPS35 was also observed in the AP-MS SILAC experiments (Dataset EV1). Lastly, pull-down experiments on HeLa cell lysates with bacterially purified GST:RAB21 [55] were performed to assess a more direct interaction between the WASH/retromer complexes and RAB21. FAM21,

Strumpellin and CAPZ, three WASH complex components, as well as VPS35, a core retromer subunit (Fig. 3D), were detected. The efficiency of the pull-down was confirmed by the presence of APPL1 (Fig. 3D). While these pull-down results do not allow concluding that RAB21 directly interacts with the WASH or retromer complexes, they nevertheless strongly suggest that RAB21 binds to these complexes either directly or indirectly. Finally, as expected from the interaction data, partial colocalization between RAB21 and the WASH and retromer complexes was observed (Fig. 3E and F), further indicating that RAB21 interacts with the WASH and retromer complexes.

RAB21 is required for complete endosomal WASH/retromer complex recruitment

To functionally assess whether RAB21 affects WASH/retromer functions, polyclonal

RAB21 knockout HeLa cell populations were generated using CRISPR/Cas9 [70]. To ensure the specificity of the assayed phenotypes, two cell populations using independent guide

RNAs targeting distinct genomic regions were generated. Thus, similar phenotypes in both cell populations would strongly argue for a specific effect. RAB21 knockout was confirmed through sequencing of the targeted genomic regions, with various indels or insertions observed for each gRNA in the polyclonal populations (Fig. EV4E). Some non-edited cells were observed in the gRNA-2 cell population compared to the gRNA-3 population (Fig.

77 78

EV4E), which correlated with RAB21 protein levels detected by Western blotting (Fig. 4A).

Importantly, as expected from previous studies [43], the number of LC3 puncta was increased in the two RAB21 knockout cell populations (Fig. EV4F). Morphological analysis of various membrane compartments did not identify major defects in compartment size and localization

(Fig. 4B), aside from transferrin uptake being slower in RAB21 KO cells (Fig. EV4G and H) in accordance with previous work [45].

RAB21 knockout led to a slight reduction in VPS35 and VPS26 protein levels, while it did not affect VPS29. On the opposite WASH, Strumpellin and CAPZ proteins levels showed a slightly increasing trend (Fig. 5A and B), indicating that RAB21 may regulate the localization or stability of these complexes. Significantly, loss of RAB21 reduced VPS35 recruitment to endosomes detected both by a decreased number of VPS35 puncta per cells and reduced colocalization between VPS35 and EEA1 (Fig. 5C to E). Decreased colocalization between SNX1 and EEA1 was also observed, although the number of endosomal SNX1 puncta was not affected by RAB21 deletion (Fig. EV5A to C), suggesting that the cargo sorting complex is more affected by the loss of RAB21 then the sorting nexin complex. WASH and FAM21 also showed, albeit to a weaker extent then VPS35, decreased endosomal colocalization with EEA1 (Fig. 5F to H and EV5D). Altogether, these results demonstrate that RAB21 is required for the full recruitment of the WASH and retromer complexes at EEA1 positive endosomes. To assess whether RAB21 was required for a specific WASH-mediated process, endosomal F-actin was monitored since it is believed to be mostly controlled by the WASH complex [27,71,72]. Importantly, endosomal F-actin was decreased in RAB21 KO cells compared to controls (Fig. 5I and J) demonstrating the

78 79 importance of RAB21 for proper WASH function. This was not a global loss of F-actin, since cortical actin was present equally in parental and RAB21 knockout cells (Fig. 5I).

RAB21 regulates trafficking of a subset of retromer cargos

The interaction and impact of RAB21 on WASH/retromer localization and actin polymerization suggests that RAB21 may regulate the trafficking of a larger proportion of cargos than initially believed [40,43]. WASH/retromer complexes regulate endosome-to-

Golgi retrograde pathways and cargo recycling between endosomes to plasma membrane

[17,73]. Thus, given that both CI-MPR and GLUT1 proteins represent two well-defined retromer cargos undergoing retrograde or endosome-to-plasma membrane sorting respectively [17,74], the localization of these two proteins was therefore assessed.

Surprisingly, in RAB21 knockout cells, no defects were observed for either cargos (Fig. 6A to C), thereby indicating that RAB21 is not required for general retromer sorting events, and consistent with the observed partial endosomal loss of WASH/retromer complexes (Fig. 5C to H). Given the recently reported heterogeneity of retromer complexes [20], it is possible that RAB21 modulates the sorting of a defined set of cargos. Indeed, VARP was recently identified as a retromer interacting protein and shown to be involved in regulating a subset of VPS35-dependent cargos. Since RAB21 KO did not impact CI-MPR or GLUT1 trafficking, the VARP-dependent cargo MCT1 was assessed. RAB21 KO notably affected

MCT1 protein levels (Fig. 6D and E). This defect is reminiscent of what was observed in

VARP-depleted cells [20], thus indicating that RAB21 modulates a subset of retromer cargos.

Interestingly, MCT1 (SLC16A1) was present in both GFP-Trap and APEX2:RAB21 datasets. Since SLC3A2, Basigin (CD147) and SLC16A3 were also observed in both datasets

79 80

(Datasets EV1 and EV2), this could indicate that their trafficking may be regulated by

RAB21, similarly to MCT1. The interaction between RAB21 and SLC3A2 (Fig. 6F) was thus first validated. Confirming the proteomic data, FLAG:RAB21 efficiently immunoprecipitated HA:SLC3A2. Given the available reagents, only SLC3A2 and Basigin trafficking was subsequently investigated. Importantly, these two cargos represent ‘direct

CIE cargos’ [75], which are characterized by their rapid sorting into tubular endosomes in

HeLa cells [76]. As a result of this feature, their tubular endosomal localization was monitored through a well-defined antibody uptake assay [77]. As expected, SLC3A2 and

Basigin both localized to endosomal tubules in control cells (Fig. 6G and H). Conversely, in

RAB21 KO cells, both failed to reach tubular endosomes and were observed in a vesicular pattern (Fig. 6G and H). Importantly, SLC3A2 presence on tubules was partially rescued by transient overexpression of FLAG:RAB21 WT (Fig. EV5E). CD44, another well-defined

‘direct CIE cargo’, which was not observed in either the AP-MS or APEX2 datasets, was also followed in order to monitor whether RAB21 only affected interacting cargos, or rather acted more generally in tubular endosome sorting events. CD44, similarly to SLC3A2 and

Basigin, failed to reach endosomal tubules in RAB21 KO cells (Fig. 6G and H). These findings thus suggest a more general role of RAB21 in mediating fast endosomal sorting of

‘direct CIE cargos’.

Finally, to further corroborate that the loss of RAB21 affects the sorting of these cargos, steady-state SLC3A2 protein levels were followed, with the prediction that the latter would decrease due to lysosomal degradation, as observed for many retromer-dependent cargos

[17]. In accordance with a sorting defect, an increased colocalization between SLC3A2 and late endosomes (RAB7) and lysosomes (LAMP1) (Fig. EV5F and G) was detected.

80 81

Moreover, SLC3A2 total protein levels were decreased in RAB21 KO cells, as assessed by both FACS and immunoblotting (Fig. EV5H and I). Altogether, these findings illustrate a new role for RAB21 in controlling endosomal sorting of direct clathrin-independent cargos.

Both WASH and retromer complexes are required for SLC3A2 and Basigin sorting and for full RAB21 activity

To confirm that SLC3A2 and Basigin sorting to endocytic tubules requires the WASH and retromer complexes, knockout HeLa cell populations for FAM21 (WASH), VPS29

(retromer) and VARP (retromer subcomplex) (Fig. 7A) were generated. Of note, both complexes were needed for full sorting of SLC3A2 and Basigin into endosomal tubules (Fig.

7B and C). Although the effect was weaker compared to RAB21 (Fig. 6H), there was a clear drop in the number of cells harboring SLC3A2- and Basigin-labeled tubules in FAM21,

VPS29 and VARP knockout cells (Fig. 7B and C). Importantly, SLC3A2 total protein levels were also downregulated in WASH and retromer knockout cells (Fig. 7A and D). This latter finding is also in accordance with previous proteomic studies [17,78] and argues for a common trafficking pathway essential for both SLC3A2 and Basigin sorting and requiring

RAB21, WASH and the retromer.

RAB21 interaction with the WASH/retromer complexes was independent of RAB21 GTP status (Fig. 3C), thus suggesting that these complexes are unlikely to act as RAB21 effectors.

Hence, it is possible that the retromer may additionally regulate RAB21 activity, given its association, through VPS29, with VARP, a RAB21 GEF [20,32,79]. This would provide a potential feedforward, or amplifying loop, to ensure concomitant RAB21 activation and retromer/WASH recruitment at specific endosomal cargos. To assess this, we monitored

81 82

RAB21 endosomal localization in VPS29 and VARP KO cells. Interestingly, colocalization between RAB21 and EEA1 or VPS26 (Fig. 7E to G) decreased in KO cells. Overall, these colocalization studies suggests that the retromer also regulates RAB21 functions, most likely through its interaction with VARP.

Discussion

Defining a global picture of RAB GTPase interactors/neighbors has been hampered by inherent limitations related to the techniques used. Through the use and comparison of AP-

MS and APEX2 proximity labeling, the present data show that APEX2:RAB proximity labeling was efficient at identifying RAB interacting/neighboring proteins ranging from

GEFs, GAPs, effectors and protein complexes. While important novel interactions were uncovered through the AP-MS technique, the APEX2 approach outperformed the former in terms of specificity and coverage. In addition, a new link between RAB21 and the WASH and retromer complexes was established from analysis of the proteomic datasets. As such,

RAB21 was found to be required for complete recruitment of the WASH and retromer complexes to endosomes and for WASH-mediated actin polymerization. Moreover, RAB21 deletion compromised the trafficking of specific clathrin-independent cargos. This impairment was also observed in retromer or WASH complex-deleted cells. In light of the above, it is proposed that RAB21 and the retromer/WASH complexes operate in a feedforward loop to ensure their respective appropriate endosomal recruitment in order to mediate efficient cargo sorting (Fig. 8).

By comparing the neighboring proteomes of three early endosomal RABs, a strong overlap between the latter was observed, in addition to specific enriched neighbors for each of these

82 83

RABs, in accordance with the concept of RAB microdomains [51,52]. These observations support the use of APEX2 as an efficient tool to identify RAB regulators/effectors. One caveat however with APEX2 is that it does not discriminate between direct or indirect interactors, nor with bystander proteins. In general, proximity labeling often leads to the identification of a large quantity of proteins [50]. Indeed, there are a large number of statistically enriched proteins for all RABs, thus rendering their follow-up study prioritization difficult. Over 1375 significant proteins were identified herein when combining four RABs.

While they most probably not all represent direct interactors or direct RAB/protein complexes, the present data indicate that they reflect the protein environment associated with each RAB and that mining through these neighbors will likely be helpful for generating new hypotheses on RAB functions. However, because of the large number of identified proteins, it will be important to combine multiple approaches or datasets to help in establishing priorities. Here, the combination of GFP:Trap and APEX2 led us to identify a novel link between RAB21 and sorting of a subset of clathrin-independent cargos.

Given the vast amount of proteins recovered by APEX2:RAB, it is important to use appropriate controls to identify biologically relevant candidates. In our hands, normalization to APEX2 and the direct comparison of related RABs yielded more valuable information than an ectopically endosomal-enriched probe. Moreover, when using a previously published

APEX2:2xFYVE dataset [59], we observed that certain RAB interactors would have been filtered-out, since they are also associated with PtdIns(3)P binding proteins (i.e. EEA1 [80]).

Another approach that could define proteins more directly associated with RABs would be to use anti-biotin immunoprecipitation of trypsin-digested APEX2:RAB lysates in order to directly map biotinylated peptides [68]. Comparing the latter over a general streptavidin

83 84 enrichment would in principle allow mapping of more direct interactions, as recently shown

[81], and would most probably significantly reduce the number of identified-proteins.

Finally, we believe that expanding the repertoire of APEX2:RAB to each cell compartment will help toward identifying specific neighbors as already performed for phosphatases [50], and as observed herein when comparing APEX2:RAB7 to the three early endosomal RABs

(Fig. 2).

Since RAB21 proximity labeling enriched WASH and retromer complex members, the functional role between RAB21 and the WASH/retromer complexes was further investigated. Importantly, results from the co-immunoprecipitation and pull-down experiments showed that RAB21 interacts with multiple endogenous WASH and retromer subunits. These findings strongly argue in favor of RAB21 binding to the WASH and retromer complexes. The fact that FAM21, Strumpellin and CAPZ were observed in pull- down studies further suggests that RAB21 directly interacts with the full WASH complex

[82]. Nevertheless, the data do not allow to firmly conclude on this latter aspect and more defined structure-function studies will be required to map the WASH interaction site, as is also the case for VPS35 pull-down. It will be furthermore important in future studies to define which WASH- or retromer-specific subunit(s) is/are bound by RAB21 and to map the binding domains. Finally, congruent with our co-immunoprecipitation and pull-down results, RAB21 depletion partially impaired endosomal recruitment of the WASH and retromer complexes and also led to a decrease in endosomal F-actin and cargo trafficking impairments (Fig. 8).

84 85

Given the observed role of RAB21 for WASH/retromer functions, one could have expected broader defects in RAB21 KO cells. No endosomal collapse, as observed in WASH or

FAM21-depleted cells [83], or ectopic endosomal tubulation was observed in RAB21 deficient cells [19,71]. This is probably due to the fact that the WASH and retromer complexes are recruited at endosomes through multiple pathways. Both SNX3 and RAB7 have been shown to mediate endosomal retromer recruitment [13,14,84], while FAM21 binding to VPS35 has been found to be an important determinant of WASH endosomal recruitment [19,25,85]. Although, recent studies have found that WASH can be recruited to endosomes independently of VPS35 [33,86]. Since only a partial reduction in retromer/WASH endosomal association was observed herein in RAB21 KO cells, a model in which RAB21 would be required for endosomal recruitment of a specific (or several)

WASH/retromer complex(es) involved in the sorting of a subset of cargos is more in accordance with the observed data. This latter view is in agreement with newer models indicating heterogeneity in retromer-sorting decisions and complexes [20,22,23]. Hence, while necessitating further confirmation, an appealing hypothesis would be that RAB21 could act in addition to RAB7 and SNX3, and directly recruit the WASH and retromer complexes to specific cargos. Another interesting possibility would be that RAB21 could function with either SNX3 or RAB7 in a co-incidence detection mechanism in order to recruit

WASH/retromer complexes at endosomes, again to a subset of cargos. In this model, RAB21 would specify the type of WASH/retromer subcomplexes involved.

Another important aspect of this study is the decrease in detectable endosomal F-actin observed in RAB21-depleted cells. The exact mechanism by which RAB21 regulates WASH actin generation is unclear, given that WASH endosomal localization was only mildly

85 86 affected in RAB21 KO cells. To draw an analogy with the WAVE complex [87,88], one could speculate that the WASH complex also requires the action of multiple inputs to drive full activation and actin polymerization. Ubiquitination was shown to regulate WASH activity [72], and it is conceivable that RAB21 could modulate this process, given its proximity to USP7 (Fig. EV4B). Alternatively, it is also possible that RAB21 could modulate endosomal F-actin independently of WASH, through recruitment of another RAB21 interacting protein. Although outside the scope of this study, it would be interesting to test whether RAB21 can modulate WASH activation directly or indirectly. In keeping with the role of RAB21 in the endosomal tubule sorting of SLC3A2 and Basigin, it is worth noting that F-actin was shown to be required for formation and sorting of ‘direct CIE cargos’ [89].

This may also explain why RAB21 was also required for CD44 sorting, even though it did not interact with CD44. Given our findings that RAB21 and the WASH/retromer complexes are required for ‘direct CIE cargos’ sorting, it could be suggested that these sorting events require F-actin generation, most probably for enrichment of cargos in tubules as previously reported for ß2AR [30] or for the formation of the tubules. Since loss of RAB21 had a stronger effect on CIE cargo tubule sorting, it is also possible that RAB21 plays a structural role by recruiting other effectors involved in the generation of these tubules, as also observed for RAB22 and RAB35 [90,91].

Of the various cargos tested herein, only ‘direct CIE cargos’ were strongly impaired along with MCT1, a recently identified VARP-dependent cargo [20]. Our data thus suggest that

MCT1 might also represent a ‘direct CIE cargo’, a possibility worth testing. The requirement of RAB21 for MCT1 trafficking also further strengthens our working model in which RAB21 would be involved in recruiting specific WASH/retromer subcomplexes to direct endosome-

86 87 to-plasma membrane cargo trafficking. These ‘specific’ WASH/retromer subcomplexes would most probably include VARP while excluding TBC1D5 since VARP and TBC1D5 binding sites on VPS29 share the same interface and are thought to be mutually exclusive

[19,32].

Another noteworthy finding from the current proteomic experiments was the specific presence of TBC1D23 in APEX2:RAB21-proximity labeling. Recent work has identified

TBC1D23 as a bridge factor for endosomal-to-Golgi carriers [92]. In this study, TBC1D23 was shown to interact directly and simultaneously with either golgine-97 or golgine-245 and the WASH complex [92]. Importantly, TBC1D23, like the WASH complex, was required for TGN46 endosome-to-Golgi trafficking. It was also proposed that another factor could act together with FAM21 to increase the specificity of captured vesicles. Given our demonstrated interaction with the WASH complex and the presence of TBC1D23 in APEX2:RAB21,

RAB21 could represent such a factor. However, we did not observe strong defects in either

CI-MPR or TGN46 trafficking in RAB21 KO cells. While this latter observation argues against RAB21 playing a predominant role in this retrieval pathway, RAB21 could nonetheless still be involved in specifying vesicle subtypes that are different from CI-MPR- and TGN46-containing vesicles.

The WASH complex has also been recently associated with the Retriever complex [33]. This

VPS29-, C16orf62-, DSCR3-containing core complex, akin to the VPS35, VPS29 and

VPS26 cargo retromer complex, functions with SNX17 to regulate the sorting of a large array of cargos. Importantly, proteomic analyses showed distinct and overlapping cargos with the retromer [33]. In the present study, C16orf62 was enriched in RAB21 proximity labeling

87 88 whereas DSCR3 and SNX17 were not, while CCC complex members were observed with variable enrichment ratios (Dataset EV2). Hence, RAB21 could possibly be involved in regulating WASH and retriever complex formation and/or sorting. Our experimental validation of RAB21 was specifically focused on the shared roles of RAB21 with the WASH and retromer complexes. In the present instance, retromer versus retriever specific functions were not discriminated due to the use of a VPS29 knockout. Given the role of VPS29 with the retriever complex, certain observed phenotypes could have been caused by a loss of retriever activity. However, our demonstration that RAB21 functionally interacts with the

WASH and retromer complexes is robust given that all knockouts shared similar SLC3A2 and Basigin trafficking impairments. Significantly, loss of VARP function, which is independent of retriever activity and associated with the retromer, shared the same RAB21 phenotypes on cargos. Importantly, endosomal c16orf62 localization was not decreased by

RAB21 deletion, but rather increased (Figure EV5J), thus suggesting that RAB21 could potentially modulate the balance between retromer and retriever complex association with endosomes, possibly through RAB21 interaction with the WASH complex or through VARP.

Again, this represents an exciting new possibility which will require further validation.

In summary, through the use of unbiased proteomics, the present study allowed uncovering novel functional roles for RAB21 in direct clathrin-independent sorting events. Results also demonstrate the robustness and ease of APEX2-mediated proximity labeling and its applicability to efficiently identify novel RAB specific functions. This approach may be further extended to temporally generate dynamic findings on RAB GTPases under various cellular stimulations, thereby leading to a better definition of RAB GTPase regulation and their association with their respective effectors.

88 89

Materials and Methods

Cell culture

HCT116 and HeLa Flp-in/T-REx cells and HeLaM cells (gift from T. Yoshimori) were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (Wisent) at 37°C and 5% CO2. FT cells were maintained under selection with

100 µg/ml Zeocin and 5 µg/ml blasticidin. FT control cells are the non-recombined parental cells and were used as control in all SILAC experiments, as performed previously [44,93].

All cells were routinely screened for mycoplasma contaminations, and experiments excluded if contamination was observed. To recombine the various GFP:RAB21 variants,

APEX2:RAB4a, 5a, 7a and APEX2:RAB21 in cells, FT cells were transfected with

JetPRIME (Polyplus), according to manufacturer’s instruction. A RAB/pOG44 DNA ratio of 1 to 10 was used, such that 0.1 µg of pGLAP:RAB were cotransfected along with 0.9 µg of pOG44 (ThermoFisher) in individual wells of a 6-well plate. Twenty-four hours following transfection, recombinant cells were selected with 5 µg/ml blasticidin and 100 µg/ml hygromycin (150 µg/ml for HeLa cells). Following selection, cells were pooled and a polyclonal population was expanded and used for all subsequent experiments. RAB21,

RAB4, RAB5, RAB7 expression was achieved through addition of doxycycline at 10 ng/ml final concentration for 24 hours prior to all experiments.

For SILAC experiments, FT cells were maintained in light DMEM (R0K0: Arg0, Sigma,

A5006; Lys0, Sigma, L5501), while GFP:RAB21 variants were grown in medium (R6K4:

Arg6, Cambridge Isotope Lab (CIL), CNM-2265; Lys4, CIL, DLM-2640) or heavy (R10K8:

Arg10, CIL, CNLM-539; Lys8, CIL, CNLM- 291) DMEM depleted of arginine and lysine

(Life Technologies A14431-01), and supplemented with 10% dialyzed fetal bovine serum

89 90

(FBS) (Invitrogen, 26400-044), called SILAC medium. Cells were passaged 5 times to ensure proper incorporation of all heavy amino acids, prior to SILAC experiments.

Generation of DNA constructs

GFP:RAB21 variants were PCR amplified from previously published pcDNA3 variants [43] and ligated using the In Fusion HD kit (Clontech) into a PCR amplified modified pGLAP1 vector [44] where GFP was placed N-terminally to the fusion protein. APEX2:RAB21-WT was generated by PCR amplification of RAB21-WT from pCDNA3-GFP:RAB21-WT and subcloned by In Fusion HD in pGLAP1-APEX2. This vector is a modified version of N- terminal pGLAP1 vector, where the GFP was replaced with Myc:APEX2 [94].

APEX2:RAB4a, APEX2:RAB5a and APEX2:RAB7a were generated by PCR amplification of RAB4a, 5a and 7a from HeLa cell cDNA. cDNA was generated with the SuperScript III

First-Strand synthesis kit (Thermo Fisher Scientific) and PCR performed with the Phusion enzyme (New England Biolabs). PCR fragments were cloned by In Fusion HD in pGLAP1-

APEX2. All clones were subsequently sequenced and validated.

Immunoprecipitations, mass spectrometry, pull-downs and immunoblots

A total of 8 x 106 cells were plated in 150 mm plates and grown for two days in appropriate

SILAC medium, followed by a 24-hour doxycycline induction (10ng/ml). Cell lysis was performed in 2.2 ml of CoIP buffer (25 mM Tris-HCl pH 7.4, 1 mM EDTA, 0.1 mM EGTA,

15 mM MgCl2, 150 mM NaCl, 2 mM Na3VO4, 10% glycerol, 1% IGEPAL CA-630, 2x protease inhibitors) per plate and cells incubated 20 minutes on ice. Lysates were cleared by centrifugation at 13,200 RPM for 10 minutes at 4°C. Cell lysates were quantified through a

BCA assay (Pierce) and equal amounts were immunoprecipitated in individual tubes with

90 91

20µl of GFP-Trap beads (Chromotek), and incubated on a rotator for 2.5 hours at 4°C. Beads were washed twice with full lysis buffer and twice again with lysis buffer lacking IGEPAL

CA-630. This was followed by five washes of 20 mM NH4HCO3. Following these washes,

GFP-Trap beads from three independent immunoprecipitations (light, medium and heavy) were mixed equally and processed for on-beads digestion and mass spectrometry analysis following these steps. Proteins were reduced 30 min with 10 mM DTT and alkylated 1 hour with 15 mM iodoacetamide. After iodoacetamide quenching with 15 mM DTT, proteins were digested overnight with 1 g trypsin. Digestion was stopped by acidification with 1% formic acid, supernatant was collected, and residual peptides eluted with 60% acetonitrile and 0.1% formic acid. Samples were then dried, resuspended in 0.1% TFA solution and desalted on a

Zip Tip. Trypsin digested peptides were loaded and separated using an Ultimate 3000 nanoLC (Thermo Fischer Scientific Inc). Ten µl of the sample (2 µg) resuspended in 1%

(v/v) formic acid was loaded onto a trap column (Acclaim PepMap100 C18 column, 0.3 mm id x 5 mm, Dionex Corporation, Sunnyvale, CA) and the peptides separated by a PepMap

C18 nano column (75 µm x 50 cm, Dionex Corporation) with a linear gradient of 5-35% solvent B (90% acetonitrile with 0.1% formic acid) over a 4 h gradient with a constant flow of 200 nl/min. Acquisition of the full scan MS survey spectra (m/z 350-1600) in profile mode was performed in a QExactive OrbiTrap (Thermo Fischer Scientific Inc) at a resolution of

70,000 using 1,000,000 ions. All unassigned charge states as well as singly, 7 and 8 charged species for the precursor ions were rejected. To improve the mass accuracy of survey scans, the lock mass option was enabled. Data acquisition was performed using Xcalibur version

2.2 SP1.48.

91 92

GST and GST:RAB21 were purified exactly as described in [95] for GST:APPL1 and pull- downs were carried out with equal amounts of GST or GST:RAB21. Pull-downs were performed as follow. 3.5x106 Hela cells were plated in 100 mm plates and 24 hours later, cells were lysed for 20 min on ice with 1 ml of MLB modified buffer (25 mM HEPES, 150 mM NaCl, 1% IGEPAL CA-630, 10% glycerol, 20 mM MgCl2, 1 mM sodium orthovanadate, 100 µM EGTA and 100 µM GTP) supplemented with protease inhibitors.

Lysates were cleared by centrifugation at 13,200 RPM for 10 min at 4°C. In the meantime,

GST and GST:RAB21 sepharose beads were washed three times in MLB modified buffer minus IGEPAL CA-630 and incubated 20 min at 4°C with rotation in modified MLB buffer without IGEPAL CA-630. Beads were then washed 3 times with MLB modified buffer and incubated with 900 µL of lysates for 1 hour at 4°C with rotation. Finally, lysates were discarded and beads were washed 3 times with MLB modified buffer containing 0.2%

IGEPAL CA-630. Media was removed and 30 µL of 2xSDS loading buffer was added to each sample.

For immunoblots, protein extracts were separated on 4-20% TGX precast gels (BioRad) and transferred on PVDF (Millipore) membranes. Antibodies used for immunoblotting were anti-

GFP (1:500, Roche #11814460001), anti-RAB21 (1:1000, Sigma #R4405 or 1:1000,

Invitrogen #PA5-34404), anti-RAB4 (1:1000, Cell Signaling #2167), anti-RAB5 (1:1000,

Cell Signaling #3547), anti-RAB7 (1:1000, Cell Signaling #9367), anti-GAPDH-HRP

(1:1000, Cell Signaling #8884), anti-Myc (1:1000, Cell Signaling #2278), anti-HA (1:1000,

Cell Signaling #3724), anti-SLC3A2 (1:800, Cell Signaling #13180), anti-Strumpellin

(1:500, Santa Cruz #377146), anti-VPS35 (1:500, Santa Cruz #374372), anti-VPS26 (1:500,

92 93

Santa Cruz #390304), anti-VPS29 (1:500, Santa Cruz #398874) anti-FAM21 and anti-

WASH (1:10 000, Gift from D. Billadeau), anti-CAPZ (1:500, Santa Cruz #374302), anti-

APPL1 (1:1000, Cell Signaling #3858), anti-VARP (1:500, Bethyl Laboratories #A302-

997A), anti-tubulin (1:2500, Sigma #T9026), streptavidin-HRP (1:1000, ThermoFisher

#N100) and anti-Rabbit and mouse-HRP (1:10 000, Jackson Laboratories #115-035-144 and

#115-035-146 respectively). Luminata-Forte (Millipore) or Clarity Max chemiluminescent substrates were used and membranes imaged on a BioRad Chemidoc XR station.

APEX2

A total of 8 x 106 HeLa cells were plated in 150 mm plates and induced 24 hours later with

10 ng/ml doxycycline for 24 hours. Freshly prepared biotin-phenol was added to full DMEM to yield a final concentration of 500 µM. After renewal of the culture medium, cells were incubated in DMEM-biotin-phenol for 30 min at 37°C. Biotinylation was achieved by adding freshly prepared H2O2 at a final concentration of 2 mM for exactly 1 min. Reactions were halted by removing the media, transferring the cells on ice and by performing five washes, each wash for 1 min, in freshly prepared quencher buffer (1x PBS containing, 10 mM sodium azide, 10 mM sodium ascorbate and 5 mM trolox(Sigma)). Cells were then processed as detailed for co-immunoprecipitations (coIPs), with the following changes. Protein lysates were incubated with 20 µl of streptavidin agarose beads (GE Lifesciences) or 20µl of Biotin

Antibody agarose beads (ImmuneChem Pharmaceuticals) for 2.5 hours at 4°C on a rotating wheel. Beads were then washed twice in coIP buffer containing 1% IGEPAL CA-630, twice in coIP buffer minus IGEPAL CA-630 and five times with 20 mM NH4HCO3. Lysates were

93 94 analyzed by immunoblots and affinity-purified biotinylated proteins were processed for

Western blots or for on-beads digestion and mass spectrometry analysis as described above.

Experimental design and statistics

For SILAC experiments, two biological repeats were performed in each cell line for each

RAB21 variant, yielding a total of 4 independent samples per variant analyzed when Hela and HCT116 data were combined. For APEX2 experiments, results were obtained from three independent biological repeats. Controls (FT) were included in all SILAC mass spectrometry experiments. Hence, a total of 6 independent FT controls were used in SILAC experiments and 3 APEX2 alone controls in APEX2 experiments. Statistical analyses of objects count or colocalization results were performed using Prism 7 software. Endosomal tubules were assessed manually. Briefly, the number of cells per field was established using DAPI staining, and cells harboring ≥2 tubules were counted as positive. An average per field was established and multiple fields of at least two independent experiments were pooled and analyzed using Prism software. For all analysis, the average number of object puncta or tubules or the average Pearson correlation per cell was tested for normality using a

D’Agostino-Pearson omnibus normality test. Samples that showed normal distribution were analyzed through unpaired t-tests, while samples showing a non-normal distribution were compared through nonparametric Mann-Whitney tests to assess the significance between the various conditions. All graphs display SEMs to assess variations within each group. For western blots quantifications, bands were quantified on the image lab software (Bio-Rad) and normalized to parental cells. One sample t-tests were performed for statistical analyses.

Mass spectrometry hit selection and network generation

94 95

After analysis by nano LC MS/MS, peptides and proteins were identified by MaxQuant version 1.5.2.8 using Uniprot (Homo sapiens, 16/07/2013, 88354 entries, Datasets EV1 and

EV2). Trypsin/P was the set enzyme, with no cleavages on arginines or lysines preceding a proline, with a maximum of two miscleavages being allowed. Mass tolerance of 7 and 20 ppm were used for precursor and fragment ions, respectively. Fixed carbomidomethyl modification on cysteine and variable oxidation on methionine and N-terminal acetylation were settled. For reliable identification, all proteins needed to complete a False Discovery

Rate (FDR) inferior to 1%: all proteins whose hits from the forward database were not 100- fold superior to hits from the reverse database were discarded. For SILAC conditions, the re- quantification option was selected. A minimum of 2 quantified peptides was settled for proteins to be considered. For a peptide, following its identification, MaxQuant used the median ratio to approximate a SILAC value for that particular peptide. Likewise, in order to establish a protein-SILAC ratio, MaxQuant used the median ratio of all identified peptides, a method shown to be appropriate for estimation of SILAC ratios [96]. To determine the

RAB21 interactome from SILAC GFP trap experiments, only proteins with an intensity

(RAB21-expressed condition)/Intensity (FT control condition) ratio > 2 in both experimental repeats were considered as potential RAB21 interactors [44,93]. In the case of APEX2 affinity purification mass spectrometry (AP-MS) experiments, the Prohit software suite [97] was used to compare all samples (APEX2, APEX2:RAB4a, APEX2:RAB5a,

APEX2:RAB7a, APEX2:RAB21 and APEX2:2xFYVE (the raw MS data for

APEX2:2xFYVE was downloaded from the ProteomeXchange Consortium). First,

SAINTexpress was run on the Crapome platform to establish high probability interactors.

Default settings were used to compare user control, in this instance, APEX2 to each

APEX2:RABs [53]. Only proteins with SAINT scores above 0.95 were selected. Thereafter,

95 96

Saint raw results were extracted and used to generate dotplots or cluster plots on a Prohits- viz platform. In order to account for experimental variations, all samples were normalized on

Prohits-viz for dotblot representations. Once RAB21 potential interactors were identified,

RAB21 protein interaction networks were generated on Cytoscape. Results were combined by cell lines (HeLa or HCT116) and once interaction networks were created, proteins were sorted based on their Genecard and Uniprot subcellular localization data. Each color corresponds to a different localization. Cytoscape platform was used to generate biological

Reactome process enrichments using Reactome plugging among specific generated networks. To compare differential protein enrichments depending on RAB21 status, SILAC

RAB21-variant mean ratios were transformed in log2(x) and 0 was imputed to non-available data. Default settings from the Perseus software [98] were used to generate hierarchical clustered heat maps.

Immunofluorescence, colocalization and transferrin or antibody uptake

A total of 70 000 HeLa or HCT116 cells were plated on glass coverslips (#1.5) and grown for 24 hours followed by GFP:RAB21 variants induction as performed above. Cells were then washed twice with 1x PBS and fixed for 15 minutes at room temperature in 4% paraformaldehyde in PBS or, depending on the antibody, in 100% methanol at -20˚C for 20 minutes. Cells were then washed 3 times with PBS for 5 min each, blocked and permeabilized for 60 min at room temperature in PBS containing 5% goat serum and 0.3% Triton X-100.

Cells were then incubated in primary antibodies overnight at 4˚C in PBS containing 1% BSA,

0.3% Triton X-100. Primary antibodies were washed three times for 5 min in PBS at room temperature and incubated at room temperature for 1 hour in secondary antibodies diluted in antibody dilution buffer. Following secondary antibody incubation, three 5 min washes in

96 97

PBS at room temperature were performed and cells were mounted in DAPI-containing mounting media (Sigma) and subsequently imaged. Antibodies used were anti-EEA1 (1:100,

Cell Signaling #3288) or (1:1000 BD Biosciences #610456), anti-SLC3A2 (1:800, Cell

Signaling #13180) or (1:400 Santa Cruz #376815), anti-Transferrin receptor (1:100, Cell

Signaling #13113), anti-RAB7 (1:250, Santa Cruz #376362), anti-LAMP1 (1:250, Santa-

Cruz #20011), anti-LC3B (1:100, Cell Signaling #3868), anti-APPL1 (1:100, Cell Signaling

#3858), streptavidin-Alexa647 (1:500, ThermoFisher #S21374), anti-FAM21 or anti-WASH

(1:1000, gift from D. Billadeau), anti-VPS26 (1:100, Abcam #23892), anti-VPS35 (Santa

Cruz, 1:100 #374372), anti-SNX1 (1:100, Abcam #ab995), Phalloidin-Alexa-488 (1:1000,

Invitrogen #A12379), anti-CIMPR (1:100, BioRad #MCA2048T), anti-TGN46 (1:100,

NovusBiological #NBP1-49643SS), anti-Glut1 (1:500, Abcam #115730), anti-MCT1

(1:100, Genetex #GTX631643) and goat anti-mouse or rabbit Alexa 488 or 546 (1:250,

ThermoFisher #A11030, A11035, A11029 and A11034). DAPI was used to unbiasedly identify fields where images were acquired.

For transferrin uptake, cells were plated at a density of 40 000 cells on (#1.5) glass coverslips and grown for 24 hours. Cells were serum starved for 30 minutes in DMEM. Alexa 647- labeled transferrin was added at 25 µg/ml and allowed to bind to the receptor for 15 minutes on ice. Cells were washed with cold PBS and incubated in FBS-supplemented DMEM for 0 to 60 minutes at 37°C. Cells were washed in PBS and fixed with 4% PFA for 15 minutes at room temperature and mounted for imaging. SLC3A2 (clone MEM-108, Biolegend), Basigin

(clone HIM6, Biolegend) and CD44 (clone BJ18, Biolegend) uptake assays were performed exactly as previously described [77].

97 98

Images for panels 1D, 4B, EV1C, EV4G and EV5F were acquired on an Olympus FV1000 confocal microscope equipped with a 60x/1.42NA plan-Apo N objective at a 1x zoom

(colocalization) or a 40x/1.3NA UPLANFLN objective (LC3 counts). Single Z-sections were acquired and acquisition settings were set to avoid pixel saturation to ensure proper colocalization and intensity quantification. Images were adjusted similarly between conditions to accurately represent the raw data. Images for panels 3E, 5C,F,I, 6A,D,G,

7B,E,F, EV5A,D,J were acquired on a ZEISS LSM880 equipped with a 40x/1.4NA plan-

Apo objective at a 1.6x zoom. For colocalization experiments, single Z-sections were acquired and gain/offset settings were similarly adjusted between conditions. For some antibodies, offset was utilized in order to remove non-specific intracellular background signal and to facilitate endosomal localization visualization. For endosomal tubule imaging, three

Z-sections were acquired at 0.36µm each and maximum intensity projections were generated.

These maximum intensity projections were used for data quantification and representation.

Finally, only linear modifications of levels were performed on Photoshop CC2017. Cropping and resolution changes to 300dpi was also performed on Photoshop for data presentation.

Image colocalization was achieved with CellProfiler [99]. Briefly, GFP:RAB21 cells were identified manually or through a primary object identification module and the correlation coefficient (Pearson) was measured on original images using the measure correlation module.

Number of object puncta per cells was also achieved using CellProfiler. For both colocalization and objects count, all measurements were exported to Prism for statistical analyses.

Generation of knockout HeLa cells

98 99

RAB21 knockout HeLa cell populations were engineered using the Crispr/Cas9 technique.

Briefly, guide RNAs were selected based on [100]. Two independent gRNAs (gRNA-2

‘AGTAAATTGGACCCAATGCA’ and gRNA-3 ‘CCACCTTGAACGAGTAGGCT’) were selected and cloned into pSPCas9(BB)-2A-Puro. A total of 1 x 106 HeLa cells were plated in

100mm dishes and transfected with individual plasmids. After 24 hours, transfected cells were selected with 1.5 µg/ml puromycin for 5 days, with media renewed every 48 hours.

Selected cells were amplified and frozen at low passages. Deletion of the RAB21 gene was verified by PCR and sequencing. Briefly, genomic DNA isolated from each gRNA-selected cell population was isolated. Amplicons encompassing predicted cut sites were amplified by

PCR (using Phusion polymerase) and ligated into pBluescript using T4 DNA ligase.

Individual bacterial clones were selected and sequenced and the representation of the cutting events are depicted in Fig. EV4E. All experiments performed on knockout cells were performed at low passages to avoid competition from potential non-targeted wild type HeLa cells.

HeLa FAM21, VPS29 and VARP knockout cell populations were performed as in [101].

Briefly, three independent gRNA cloned into the PX330A plasmid were cotransfected with the pEGFP-Puro plasmid (addgene #45561) at a ratio of 1:1:1:0.25. Transfections were performed as detailed above. Twenty-four hours after the transfections, cells were selected for 24h with 2µg/ml puromycin. Cells were amplified and used at low passages for all experiments. gRNA sequences for VPS29 and FAM21 have been published in [101], while gRNA for VARP were AAGAGCCACTCACGTCCTCG,

CAGCACGGCTACTGACTATG and TAAATACGTGGGGACCATGG. Knockout efficiencies were monitored by Western blot.

99 100

FACS analysis

A total of 1 x 106 HeLa parental, gRNA KO 2 and 3 cells were trypsinized, collected and washed with cold PBS. Cells were fixed 15 min with PFA 4% and permeabilized with methanol 90%, for 30 min. Cells were labeled (or not) with SLC3A2 rabbit antibody (dilution

1:800, Cell Signaling #13180) in PBS/BSA 0.5% incubation buffer for one hour. After two

PBS washes, cells were incubated 30 min in secondary Alexa fluor 488 antibody (dilution

1:250) diluted in incubation buffer. For each condition, Alexa fluor 488 signal intensity was measured with a BD LSR Fortessa cytometer. Acquisition was made using BDFACSDIVA software, and graphics and analysis using Flowing software 2.5.1. Mean 488 intensity signal ratios from SLC3A2 labeled cells versus non-labeled cells were used to compare SLC3A2 expression levels between the different cell lines. Signal intensity distributions from each cell line were overlapped on the same frequency histogram. In order to optimize visualization, data were submitted to virtual gain, such that f(x)=60x+15000, with x being the intensity signal. The displayed diagram (Fig. EV5H) reflects one representative experiment from the three performed repeats and represents 10 000 counted cells per cell line.

Data Availability

The SILAC and APEX2 mass spectrometry proteomic data from this publication have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository

(https://www.ebi.ac.uk/pride/archive/) and assigned the dataset identifier PXD010950.

Author contributions

100 101

TDO, AL, RL, ML, DJ, LL and SJ performed experiments. CN quantified colocalization data. FMB assisted and helped with all mass spectrometry analyses. FS provided multiple reagents and suggestions. SJ and TDO designed experiments. SJ supervised students and wrote the manuscript.

Conflict of Interest

The authors declare no conflict of interest.

101 102

Acknowledgments

We thank M.L. Dubois and D. Lévesque for helpful technical assistance with mass spectrometry and SILAC experiments, V. Delcourt for help with Perseus, C. Lavoie for generously providing multiple antibodies, M. Catala for help with FACS analysis, D.

Billadeau for kindly providing the FAM21 and WASH antibodies and Amy Kiger for support in establishing tools for the current study. We also thank all members of the Jean laboratory for insightful comments during the course of this study. We thank the Proteomic platform at the Université de Sherbrooke for proteomic services and the Photonic microscopy platform for confocal use. Steve Jean and François-Michel Boisvert are members of the FRQS-Funded

Centre de Recherche du CHUS. Steve Jean is a recipient of a Research Chair from the Centre de recherche médicale de l’Université de Sherbrooke (CRMUS). This research was supported by operating grants from the Cancer Research Society (CRS), from the Natural

Sciences and Engineering Research Council of Canada (NSERC) and the Canadian Institutes of Health Research (CIHR) and by junior faculty salary awards from Canadian Institutes of

Health Research (CIHR) and Fonds de Recherche du Québec - Santé (FRQS) to S.J..

102 103

References

1. Stenmark H (2009) Rab GTPases as coordinators of vesicle traffic. Nat Rev Mol Cell Biol 10: 513–525. 2. Nicot A-S, Laporte J (2008) Endosomal phosphoinositides and human diseases. Traffic 9: 1240–1249. 3. Amoasii L, Hnia K, Laporte J (2012) Myotubularin phosphoinositide phosphatases in human diseases. Curr Top Microbiol Immunol 362: 209–233. 4. Santiago-Tirado FH, Bretscher A (2011) Membrane-trafficking sorting hubs: cooperation between PI4P and small GTPases at the trans-Golgi network. Trends Cell Biol 21: 515–525. 5. Jean S, Kiger AA (2012) Coordination between RAB GTPase and phosphoinositide regulation and functions. Nat Rev Mol Cell Biol 13: 463–470. 6. Rojas AM, Fuentes G, Rausell A, Valencia A (2012) Evolution: The Ras protein superfamily: Evolutionary tree and role of conserved amino acids. The Journal of Cell Biology 196: 189–201. 7. Hutagalung AH, Novick PJ (2011) Role of Rab GTPases in membrane traffic and cell physiology. Physiol Rev 91: 119–149. 8. Barr F, Lambright DG (2010) Rab GEFs and GAPs. Curr Opin Cell Biol 22: 461– 470. 9. Itoh T, Satoh M, Kanno E, Fukuda M (2006) Screening for target Rabs of TBC (Tre- 2/Bub2/Cdc16) domain-containing proteins based on their Rab-binding activity. Genes to Cells 11: 1023–1037. 10. Barr F, Lambright DG (2010) Rab GEFs and GAPs. Curr Opin Cell Biol 1–10. 11. Barr FA (2013) Rab GTPases and membrane identity: Causal or inconsequential? The Journal of Cell Biology 202: 191–199. 12. Grosshans BL, Ortiz D, Novick P (2006) Rabs and their effectors: achieving specificity in membrane traffic. Proc Natl Acad Sci USA 103: 11821–11827. 13. Rojas R, Van Vlijmen T, Mardones GA, Prabhu Y, Rojas AL, Mohammed S, Heck AJ, Raposo G, Van Der Sluijs P, Bonifacino JS (2008) Regulation of retromer recruitment to endosomes by sequential action of Rab5 and Rab7. The Journal of Cell Biology 183: 513–526.

103 104

14. Seaman MNJ, Harbour ME, Tattersall D, Read E, Bright N (2009) Membrane recruitment of the cargo-selective retromer subcomplex is catalysed by the small GTPase Rab7 and inhibited by the Rab-GAP TBC1D5. Journal of Cell Science 122: 2371–2382. 15. Seaman MNJ, Marcusson EG, Cereghino JL, Emr SD (1997) Endosome to Golgi Retrieval of the Vacuolar Protein Sorting Receptor, Vps10p, Requires the Function of the VPS29, VPS30, and VPS35Gene Products. The Journal of Cell Biology 137: 79–92. 16. Burd C, Cullen PJ (2014) Retromer: A Master Conductor of Endosome Sorting. Cold Spring Harbor Perspectives in Biology 6: a016774–a016774. 17. Steinberg F, Gallon M, Winfield M, Thomas EC, Bell AJ, Heesom KJ, Tavaré JM, Cullen PJ (2013) A global analysis of SNX27–retromer assembly and cargo specificity reveals a function in glucose and metal ion transport. Nat Cell Biol 15: 461–471. 18. Seaman MNJ (2012) The retromer complex - endosomal protein recycling and beyond. Journal of Cell Science 125: 4693–4702. 19. Harbour ME, Breusegem SYA, Antrobus R, Freeman C, Reid E, Seaman MNJ (2010) The cargo-selective retromer complex is a recruiting hub for protein complexes that regulate endosomal tubule dynamics. Journal of Cell Science 123: 3703–3717. 20. Mcgough IJ, Steinberg F, Gallon M, Yatsu A, Ohbayashi N, Heesom KJ, Fukuda M, Cullen PJ (2014) Identification of molecular heterogeneity in SNX27-retromer- mediated endosome-to-plasma-membrane recycling. Journal of Cell Science 127: 4940–4953. 21. Zhong Q, Watson MJ, Lazar CS, Hounslow AM, Waltho JP, Gill GN (2005) Determinants of the endosomal localization of sorting nexin 1. Mol Biol Cell 16: 2049–2057. 22. Kvainickas A, Jimenez Orgaz A, Nägele H, Hu Z, Dengjel J, Steinberg F (2017) Cargo-selective SNX-BAR proteins mediate retromer trimer independent retrograde transport. The Journal of Cell Biology 216: 3677–3693. 23. Simonetti B, Danson CM, Heesom KJ, Cullen PJ (2017) Sequence-dependent cargo

104 105

recognition by SNX-BARs mediates retromer-independent transport of CI-MPR. The Journal of Cell Biology 216: 3695–3712. 24. Mcgough IJ, Cullen PJ (2011) Recent Advances in Retromer Biology. Traffic 12: 963–971. 25. Helfer E, Harbour ME, Henriot V, Lakisic G, Sousa-Blin C, Volceanov L, Seaman MNJ, Gautreau A (2013) Endosomal recruitment of the WASH complex: Active sequences and mutations impairing interaction with the retromer. Biology of the Cell 105: 191–207. 26. Zavodszky E, Seaman MNJ, Moreau K, Jimenez-Sanchez M, Breusegem SY, Harbour ME, Rubinsztein DC (2014) Mutation in VPS35 associated with Parkinson's disease impairs WASH complex association and inhibits autophagy. Nat Comms 5: 3828. 27. Gomez TS, Billadeau DD (2009) A FAM21-containing WASH complex regulates retromer-dependent sorting. Developmental Cell 17: 699–711. 28. Derivery E, Sousa C, Gautier JJ, Lombard B, Loew D, Gautreau A (2009) The Arp2/3 Activator WASH Controls the Fissionof Endosomes through a Large Multiprotein Complex. Developmental Cell 17: 712–723. 29. Seaman MNJ, Gautreau A, Billadeau DD (2013) Retromer-mediated endosomal protein sorting: all WASHed up! Trends Cell Biol 23: 522–528. 30. Temkin P, Lauffer B, Jäger S, Cimermancic P, Krogan NJ, Zastrow von M (2011) SNX27 mediates retromer tubule entry and endosome-to-plasma membrane trafficking of signalling receptors. Nat Cell Biol 13: 715–721. 31. Steinberg F, Heesom KJ, Bass MD, Cullen PJ (2012) SNX17 protects integrins from degradation by sorting between lysosomal and recycling pathways. The Journal of Cell Biology 197: 219–230. 32. Hesketh GG, Pérez-Dorado I, Jackson LP, Wartosch L, Schäfer IB, Gray SR, McCoy AJ, Zeldin OB, Garman EF, Harbour ME, et al. (2014) VARP Is Recruited on to Endosomes by Direct Interaction with Retromer, Where Together They Function in Export to the Cell Surface. Developmental Cell 29: 591–606. 33. McNally KE, Faulkner R, Steinberg F, Gallon M, Ghai R, Pim D, Langton P, Pearson N, Danson CM, Nägele H, et al. (2017) Retriever is a multiprotein complex for

105 106

retromer-independent endosomal cargo recycling. Nat Cell Biol 19: 1214–1225. 34. Christoforidis S, Miaczynska M, Ashman K, Wilm M, Zhao L, Yip SC, Waterfield MD, Backer JM, Zerial M (1999) Phosphatidylinositol-3-OH kinases are Rab5 effectors. Nat Cell Biol 1: 249–252. 35. Christoforidis S (2000) Purification and Identification of Novel Rab Effectors Using Affinity Chromatography. Methods 20: 403–410. 36. Fukuda M (2010) How can mammalian Rab small GTPases be comprehensively analyzed?: Development of new tools to comprehensively analyze mammalian Rabs in membrane traffic. Histol Histopathol 25: 1473–1480. 37. Gillingham AK, Sinka R, Torres IL, Lilley KS, Munro S (2014) Toward a Comprehensive Map of the Effectors of Rab GTPases. Developmental Cell 31: 358– 373. 38. Christoforidis S, Zerial M (2000) Purification and Identification of Novel Rab Effectors Using Affinity Chromatography. Methods 20: 403–410. 39. Huttlin EL, Bruckner RJ, Paulo JA, Cannon JR, Ting L, Baltier K, Colby G, Gebreab F, Gygi MP, Parzen H, et al. (2017) Architecture of the human interactome defines protein communities and disease networks. Nature 545: 505–509. 40. Alanko J, Mai A, Jacquemet G, Schauer K, Kaukonen R, Saari M, Goud B, Ivaska J (2015) Integrin endosomal signalling suppresses anoikis. Nat Cell Biol. 41. Pellinen T, Tuomi S, Arjonen A, Wolf M, Edgren H, Meyer H, Grosse R, Kitzing T, Rantala JK, Kallioniemi O (2008) Integrin Trafficking Regulated by Rab21 Is Necessary for Cytokinesis. Developmental Cell 15: 371–385. 42. Burgo A, Sotirakis E, Simmler M-C, Verraes A, Chamot C, Simpson JC, Lanzetti L, Proux-Gillardeaux VER, Galli T (2009) Role of Varp, a Rab21 exchange factor and TI-VAMP/VAMP7 partner, in neurite growth. EMBO Rep 10: 1117–1124. 43. Jean S, Cox S, Nassari S, Kiger AA (2015) Starvation-induced MTMR13 and RAB21 activity regulates VAMP8 to promote autophagosome-lysosome fusion. EMBO Rep 16: 297–311. 44. Drissi R, Dubois M-L, Douziech M, Boisvert F-M (2015) Quantitative Proteomics Reveals Dynamic Interactions of the Minichromosome Maintenance Complex (MCM) in the Cellular Response to Etoposide Induced DNA Damage. Molecular &

106 107

Cellular Proteomics 14: 2002–2013. 45. Simpson JC, Griffiths G, Wessling-Resnick M, Fransen JAM, Bennett H, Jones AT (2004) A role for the small GTPase Rab21 in the early endocytic pathway. Journal of Cell Science 117: 6297–6311. 46. Yuan Q, Ren C, Xu W, Petri B, Zhang J, Zhang Y, Kubes P, Wu D, Tang W (2017) PKN1 Directs Polarized RAB21 Vesicle Trafficking via RPH3A and Is Important for Neutrophil Adhesion and Ischemia-Reperfusion Injury. CellReports 19: 2586– 2597. 47. Martell JD, Deerinck TJ, Sancak Y, Poulos TL, Mootha VK, Sosinsky GE, Ellisman MH, Ting AY (2012) engineered ascorbate peroxidase as a genetically encoded reporter for electron microscopy. Nat Biotechnol 1–9. 48. Lee S-Y, Kang M-G, Shin S, Kwak C, Kwon T, Seo JK, Kim J-S, Rhee H-W (2017) Architecture Mapping of the Inner Mitochondrial Membrane Proteome by Chemical Tools in Live Cells. J Am Chem Soc 139: 3651–3662. 49. Lee S-Y, Kang M-G, Park J-S, Lee G, Ting AY, Rhee H-W (2016) APEX Fingerprinting Reveals the Subcellular Localization of Proteins of Interest. CellReports 1–12. 50. St-Denis N, Gupta GD, Lin Z-Y, Gonzalez-Badillo B, Veri AO, Knight JDR, Rajendran D, Couzens AL, Currie KW, Tkach JM, et al. (2016) Phenotypic and Interaction Profiling of the Human Phosphatases Identifies Diverse Mitotic Regulators. CellReports 17: 2488–2501. 51. Sönnichsen B, De Renzis S, Nielsen E, Rietdorf J, Zerial M (2000) Distinct membrane domains on endosomes in the recycling pathway visualized by multicolor imaging of Rab4, Rab5, and Rab11. The Journal of Cell Biology 149: 901–914. 52. De Renzis S, Sönnichsen B, Zerial M (2002) Divalent Rab effectors regulate the sub- compartmental organization and sorting of early endosomes. Nat Cell Biol 4: 124– 133. 53. Choi H, Larsen B, Lin Z-Y, Breitkreutz A, Mellacheruvu D, Fermin D, Qin ZS, Tyers M, Gingras A-C, Nesvizhskii AI (2010) SAINT: probabilistic scoring of affinity purification–mass spectrometry data. Nat Methods 8: 70–73. 54. Knight JDR, Choi H, Gupta GD, Pelletier L, Raught B, Nesvizhskii AI, Gingras A-

107 108

C (2017) ProHits-viz: a suite of web tools for visualizing interaction proteomics data. Nat Methods 14: 645–646. 55. Jean S, Cox S, Schmidt EJ, Robinson FL, Kiger A (2012) Sbf/MTMR13 coordinates PI(3)P and Rab21 regulation in endocytic control of cellular remodeling. Mol Biol Cell 23: 2723–2740. 56. De Franceschi N, Hamidi H, Alanko J, Sahgal P, Ivaska J (2015) Integrin traffic - the update. Journal of Cell Science 128: 839–852. 57. Chotard L, Mishra AK, Sylvain M-A, Tuck S, Lambright DG, Rocheleau CE (2010) TBC-2 regulates RAB-5/RAB-7-mediated endosomal trafficking in Caenorhabditis elegans. Mol Biol Cell 21: 2285–2296. 58. Fuchs E, Haas AK, Spooner RA, Yoshimura S-I, Lord JM, Barr FA (2007) Specific Rab GTPase-activating proteins define the Shiga toxin and epidermal growth factor uptake pathways. The Journal of Cell Biology 177: 1133–1143. 59. Lobingier BT, Hüttenhain R, Eichel K, Miller KB, Ting AY, Zastrow von M, Krogan NJ (2017) An Approach to Spatiotemporally Resolve Protein Interaction Networks in Living Cells. Cell 169: 350–360.e12. 60. Priya A, Kalaidzidis IV, Kalaidzidis Y, Lambright D, Datta S (2014) Molecular Insights into Rab7-Mediated Endosomal Recruitment of Core Retromer: Deciphering the Role of Vps26 and Vps35. Traffic 16: 68–84. 61. Schindler C, Chen Y, Pu J, Guo X, Bonifacino JS (2015) EARP is a multisubunit tethering complex involved in endocytic recycling. Nat Cell Biol 17: 639–650. 62. Gershlick DC, Schindler C, Chen Y, Bonifacino JS (2016) TSSC1 is novel component of the endosomal retrieval machinery. Mol Biol Cell 27: 2867–2878. 63. Rottner K, Hänisch J, Campellone KG (2010) WASH, WHAMM and JMY: regulation of Arp2/3 complex and beyond. Trends Cell Biol 20: 650–661. 64. Shin H-W, Hayashi M, Christoforidis S, Lacas-Gervais S, Hoepfner S, Wenk MR, Modregger J, Uttenweiler-Joseph S, Wilm M, Nystuen A, et al. (2005) An enzymatic cascade of Rab5 effectors regulates phosphoinositide turnover in the endocytic pathway. The Journal of Cell Biology 170: 607–618. 65. Muhammad E, Levitas A, Singh SR, Braiman A, Ofir R, Etzion S, Sheffield VC, Etzion Y, Carrier L, Parvari R (2015) PLEKHM2mutation leads to abnormal

108 109

localization of lysosomes, impaired autophagy flux and associates with recessive dilated cardiomyopathy and left ventricular noncompaction. Human Molecular Genetics 24: 7227–7240. 66. Shin C-S, Mishra P, Watrous JD, Carelli V, Aurelio MDR, Jain M, Chan DC (2017) The glutamate/cystine xCT antiporterantagonizes glutamine metabolismand reduces nutrient flexibility. Nat Comms 8: 1–11. 67. Hao Y-H, Fountain MD Jr, Tacer KF, Xia F, Bi W, Kang S-HL, Patel A, Rosenfeld JA, Le Caignec C, Isidor B, et al. (2015) USP7 Acts as a Molecular Rheostat to Promote WASH-Dependent Endosomal Protein Recycling and Is Mutated in a Human Neurodevelopmental Disorder. Molecular Cell 59: 956–969. 68. Udeshi ND, Pedram K, Svinkina T, Fereshetian S, Myers SA, Aygun O, Krug K, Clauser K, Ryan D, Ast T, et al. (2017) Antibodies to biotin enable large-scale detection of biotinylation sites on proteins. Nat Methods 14: 1167–1170. 69. Pellinen T, Arjonen A, Vuoriluoto K, Kallio K, Fransen JAM, Ivaska J (2006) Small GTPase Rab21 regulates cell adhesion and controls endosomal traffic of beta1- integrins. The Journal of Cell Biology 173: 767–780. 70. Ran FA, Hsu PD, Wright J, Agarwala V, Scott DA, Zhang F (2013) Genome engineering using the CRISPR-Cas9 system. Nat Protoc 8: 2281–2308. 71. Derivery E, Sousa C, Gautier JJ, Lombard B, Loew D, Gautreau A (2009) The Arp2/3 Activator WASH Controls the Fission of Endosomes through a Large Multiprotein Complex. Developmental Cell 17: 712–723. 72. Hao Y-H, Doyle JM, Ramanathan S, Gomez TS, Jia D, Xu M, Chen ZJ, Billadeau DD, Rosen MK, Potts PR (2013) Regulation of WASH-Dependent Actin Polymerization and Protein Trafficking by Ubiquitination. Cell 152: 1051–1064. 73. Seaman MNJ, Gautreau A, Billadeau DD (2013) Retromer-mediated endosomal protein sorting: all WASHed up! Trends Cell Biol 23: 522–528. 74. Seaman MNJ (2009) Enhanced SnapShot: Endosome-to-Golgi Retrieval. Cell 139: 1198–1198.e1. 75. Maldonado-Báez L, Williamson C, Donaldson JG (2013) Clathrin-independent endocytosis: A cargo-centric view. Exp Cell Res 319: 2759–2769. 76. Eyster CA, Higginson JD, Huebner R, Porat-Shliom N, Weigert R, Wu WW, Shen

109 110

R-F, Donaldson JG (2009) Discovery of New Cargo Proteins that Enter Cells through Clathrin-Independent Endocytosis. Traffic 10: 590–599. 77. Maldonado-Báez L, Cole NB, Krämer H, Donaldson JG (2013) Microtubule- dependent endosomal sorting of clathrin-independent cargo by Hook1. The Journal of Cell Biology 201: 233–247. 78. Kvainickas A, Orgaz AJ, Nägele H, Diedrich B, Heesom KJ, Dengjel J, Cullen PJ, Steinberg F (2017) Retromer- and WASH-dependent sorting of nutrient transporters requires a multivalent interaction network with ANKRD50. Journal of Cell Science 130: 382–395. 79. Zhang X (2006) Varp is a Rab21 guanine nucleotide exchange factor and regulates endosome dynamics. Journal of Cell Science 119: 1053–1062. 80. Simonsen A, Lippé R, Christoforidis S, Gaullier JM, Brech A, Callaghan J, Toh BH, Murphy C, Zerial M, Stenmark H (1998) EEA1 links PI(3)K function to Rab5 regulation of endosome fusion. Nature 394: 494–498. 81. Kim DI, Cutler JA, Na CH, Reckel S, Renuse S, Madugundu AK, Tahir R, Goldschmidt HL, Reddy KL, Huganir RL, et al. (2018) BioSITe: A Method for Direct Detection and Quantitation of Site-Specific Biotinylation. J Proteome Res 17: 759–769. 82. Derivery E, Gautreau A (2010) Assaying WAVE and WASH Complex Constitutive Activities Toward the Arp2/3 Complex. Elsevier Inc. 83. Gomez TS, Gorman JA, Artal-Martinez De Narvajas A, Koenig AO, Billadeau DD (2012) Trafficking defects in WASH-knockout fibroblasts originate from collapsed endosomal and lysosomal networks. Mol Biol Cell 23: 3215–3228. 84. Harterink M, Port F, Lorenowicz MJ, Mcgough IJ, Silhankova M, Betist MC, van Weering JRT, Heesbeen RGHPV, Middelkoop TC, Basler K, et al. (2011) A SNX3- dependent retromer pathway mediates retrograde transport of the Wnt sorting receptor Wntless and is required for Wnt secretion. Nat Cell Biol 13: 914–923. 85. Jia D, Gomez TS, Billadeau DD, Rosen MK (2012) Multiple repeat elements within the FAM21 tail link the WASH actin regulatory complex to the retromer. Mol Biol Cell 23: 2352–2361. 86. Buckley CM, Gopaldass N, Bosmani C, Johnston SA, Soldati T, Insall RH, King JS

110 111

(2016) WASH drives early recycling from macropinosomes and phagosomes to maintain surface phagocytic receptors. Proc Natl Acad Sci USA 113: E5906–E5915. 87. Koronakis V, Hume PJ, Humphreys D, Liu T, Hørning O, Jensen ON, McGhie EJ (2011) WAVE regulatory complex activation by cooperating GTPases Arf and Rac1. Proc Natl Acad Sci USA 108: 14449–14454. 88. Lebensohn AM, Kirschner MW (2009) Activation of the WAVE Complex by Coincident Signals Controls Actin Assembly. Molecular Cell 36: 512–524. 89. Donaldson JG, Johnson DL, Dutta D (2016) Rab and Arf G proteins in endosomal trafficking and cell surface homeostasis. Small GTPases 7: 247–251. 90. Maldonado-Baez L, Cole NB, Kramer H, Donaldson JG (2013) Microtubule- dependent endosomal sorting of clathrin-independent cargo by Hook1. The Journal of Cell Biology 201: 233–247. 91. Dutta D, Donaldson JG (2015) Sorting of Clathrin-Independent Cargo Proteins Depends on Rab35 Delivered by Clathrin-Mediated Endocytosis. Traffic 16: 994– 1009. 92. Shin JJH, Gillingham AK, Begum F, Chadwick J, Munro S (2017) TBC1D23 is a bridging factor for endosomal vesicle capture by golgins at the trans-Golgi. Nat Cell Biol 19: 1424–1432. 93. Dubois M-L, Bastin C, Lévesque D, Boisvert F-M (2016) Comprehensive Characterization of Minichromosome Maintenance Complex (MCM) Protein Interactions Using Affinity and Proximity Purifications Coupled to Mass Spectrometry. J Proteome Res 15: 2924–2934. 94. Lam SS, Martell JD, Kamer KJ, Deerinck TJ, Ellisman MH, Mootha VK, Ting AY (2014) Directed evolution of APEX2 for electron microscopy and proximity labeling. Nat Methods 12: 51–54. 95. Jean S, Kiger AA (2016) RAB21 Activity Assay Using GST-fused APPL1. Bio Protoc 6:. 96. Guan X, Rastogi N, Parthun MR, Freitas MA (2014) SILAC Peptide Ratio Calculator: A Tool for SILAC Quantitation of Peptides and Post-Translational Modifications. J Proteome Res 13: 506–516. 97. Liu G, Zhang J, Larsen B, Stark C, Breitkreutz A, Lin Z-Y, Breitkreutz B-J, Ding Y,

111 112

Colwill K, Pasculescu A, et al. (2010) ProHits: integrated software for mass spectrometry– based interaction proteomics. Nat Biotechnol 28: 1015–1017. 98. Tyanova S, Temu T, Sinitcyn P, Carlson A, Hein MY, Geiger T, Mann M, Cox J (2016) The Perseus computational platform for comprehensive analysis of (prote)omics data. Nat Methods 13: 731–740. 99. Carpenter AE, Jones TR, Lamprecht MR, Clarke C, Kang IH, Friman O, Guertin DA, Chang JH, Lindquist RA, Moffat J, et al. (2006) CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol 7: R100. 100. Sullender M, Hegde M, Vaimberg EW, Donovan KF, Smith I, Tothova Z, Wilen C, Orchard R, Virgin HW, Doench JG, et al. (2016) Optimized sgrNA design to maximize activity and minimize off-target effects of crisPr-cas9. Nat Biotechnol 34: 184–191. 101. Jimenez Orgaz A, Kvainickas A, Nägele H, Denner J, Eimer S, Dengjel J, Steinberg F (2018) Control of RAB7 activity and localization through the retromer‐TBC1D5 complex enables RAB7‐dependent mitophagy. EMBO J 37: 235–254.

112 113

Figure legends

Figure 1: APEX2:RAB expression lead to endosomal biotinylation. (A) Diagram representing APEX2:RAB-mediated endosomal biotinylation of endogenous proteins. (B)

Illustrative representation of APEX2:RAB endosomal microdomains. APEX2 only or

APEX2:RAB were found to biotinylate endogenous proteins. (C) Streptavidin Western blotting of total biotinylated proteins in APEX2:RAB or APEX2 only Flp-In/T-REx HeLa cells. (D) APEX2:RAB biotinylated proteins (streptavidin) are partially colocalized with

EEA1 in HeLa cells, n=2 independent experiments. Scale bars: 10µm or 5µm in enlarged views.

Figure 2: Comparison of APEX2:RAB21-, 5-, 4- and 7-generated proteomes. (A) VENN diagram highlighting the number of interactors with each RAB and their overlap. (B)

Reactome analysis of all APEX2:RAB21, APEX2:RAB5, APEX2:RAB4 and APEX2:RAB7 interactors/neighbors. (C) ProHits-viz generated dotblots of RAB effectors, GEFs, GAPs and sorting complexes. Green, blue and beige shadings along with RAB names refer to previously identified interactions between RABs and the depicted proteins.

Figure 3: RAB21 interacts and colocalizes with the WASH and retromer complexes.

Strumpellin and VPS35 are in close proximity to RAB5 and RAB21. (A) Anti-biotin immunoprecipitation and immunoblot of endogenous Strumpellin and VPS35. Lysates correspond to 1% of input, n=3 independent experiments. RAB21 can be seen interacting with various WASH and retromer complexes subunits. GFP trap IP of WT, Q78L and T33N

RAB21 variants in HeLa cells followed by GFP immunoblot and endogenous (B) VPS26 immunoblot and (C) FAM21, Strumpellin and VPS35 immunoblots. Lysates correspond to

113 114

5% of input, n≥ 3 independent experiments. RAB21 actively pull-downs WASH and retromer complex subunits (D) Bacterially-purified and GTP-loaded GST:RAB21 pull-down of HeLa cell lysates followed by Strumpellin, FAM21, CAPZ and VPS35 immunoblots.

Ponceau staining reveals the purity of the GST and GST:RAB21 used for the pull-downs. n=3 independent experiments. (E) RAB21 colocalizes with the WASH and retromer complexes. Transiently expressing GFP:RAB21 HeLa cells were fixed and stained for endogenous EEA1, FAM21, WASH and VPS26. Boxed region is magnified and single channels are depicted. Scale bars: 10µm or 5µm in enlarged view. (F) Per cell Pearson

Correlation between RAB21 and the various markers. Error bars are SEM, n=2 independent experiments.

Figure 4: Characterization of RAB21 knockout HeLa cells. (A) Low RAB21 expression levels in two independent RAB21 KO cell populations. Immunoblot analysis of endogenous

RAB21, RAB5, RAB7 and GAPDH. (B) Endo-lysosomal compartments are unaffected by the loss of RAB21. Immunofluorescences of APPL1, TfR, RAB7 and LAMP1 in parental

HeLa cells and in the two RAB21 knockout cell populations. Scale bars :10µm, n=3 independent experiments.

Figure 5: RAB21 modulates WASH and retromer endosomal recruitment and WASH activity. (A) The stability of the retromer cargo sorting complex is affected in RAB21 KO cells. Immunoblot analysis of VPS35, VPS26, VPS29, GAPDH and RAB21 in parental and the two RAB21 KO cell populations. Ratio of VPS35, VPS26 and VPS29 integrated densities to GAPDH for four independent experiments; SEM. * represents p <0.05 (B) WASH

114 115 complex proteins are not affected by RAB21 deletion. Immunoblot of Strumpellin, WASH,

FAM21, CAPZ, GAPDH and RAB21. Ratio of WASH, FAM21, Strumpellin and CAPZ integrated densities to GAPDH for four independent experiments; SEM. (C-E) RAB21 knockout decreases VPS35 localization at endosomes. (C) Immunofluorescence of endogenous VPS35 and EEA1 in RAB21-depleted cells. Boxed region is magnified and single channel images are depicted. (D) Per cell Pearson Correlation between VPS35 and

EEA1; SEM, n=3 independent experiments. (E) Average number of VPS35 puncta per cell;

SEM, n=3 independent experiments. (F-H) RAB21 loss impairs endosomal recruitment of the WASH complex. (F) Immunofluorescence of endogenous WASH and EEA1 in RAB21- depleted cells. Boxed region is magnified and single channel images are depicted. (G) Per cell Pearson Correlation between VPS35 and EEA1 and (H) FAM21 and EEA1; SEM, n=2 independent experiments. (I-J) RAB21 knockout reduces endosomal F-actin levels. (I)

Immunofluorescence of endogenous F-actin using Alexa-488 conjugated phalloidin and

EEA1 in RAB21-depleted cells. Boxed region is magnified and single channel images are depicted. (J) Per cell Pearson Correlation between internal F-actin and EEA1; SEM, n=3 independent experiments.

Data information: In (C, F, I) scale bars represent 10µm or 2.5µm in enlarged views.

Statistical tests used: (A and B) – One sample t-tests, (D) – Unpaired t-tests, (E and J) –

Mann-Whitney tests. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats.

Figure 6: RAB21 is required for endosomal sorting of specific cargo types. (A-C) RAB21 is not required for CI-MPR or GLUT1 trafficking. (A) Immunofluorescence of endogenous

115 116

CI-MPR and TGN46 or GLUT1 with LAMP1 in RAB21-knockout cells. Boxed region is magnified. (B) Per cell Pearson Correlation between CI-MPR and TGN46 or (C) GLUT1 and LAMP1; SEM, n=3 independent experiments. MCT1 trafficking requires RAB21. (D)

Immunofluorescence of endogenous MCT1 in wild type or RAB21-depleted cells. (E) Per cell integrated MCT1 intensity (in Relative Fluorescence Unit); SEM, n=2 independent experiments. (F) RAB21 interacts with SLC3A2. FLAG immunoprecipitation of

FLAG:RAB21 and immunoblot of co-expressed SLC3A2:HA. Lysates correspond to 5% input, n=3 independent experiments. (G-H) RAB21 is required for appropriate SLC3A2,

Basigin and CD44 trafficking. (G) Antibody-uptake assays in parental or knockout RAB21 cells. Cells were stained with an Alexa488 conjugated anti-mouse antibody. Arrowheads point to endosomal tubules. Note that non-internalized antibodies were removed by an acid wash, which was not fully efficient for CD147 and CD44, explaining plasma membrane labeling. (H) Percentage of cells with tubules in parental or RAB21 knockout cells; SEM, n=3 independent experiments.

Data information: In (A), scale bars represent 20µm or 5µm in enlarged views for CI-MPR, while they represent 10µm in enlarged views for GLUT1. (D and G) scale bars represent

20µm or 5µm in enlarged views. Statistical tests used: (B) – Unpaired t-tests, (C and H) -

Mann-Whitney tests. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats

Figure 7: WASH and retromer complexes are required for SLC3A2 and Basigin sorting, and for full RAB21 activation and endosomal localization. (A) Validation of

WASH and retromer knockouts and SLC3A2 protein levels. Immunoblots of parental or

FAM21, VARP and VPS29 knockout cell populations. (B) Antibody-uptake assays in

116 117 parental or knockout FAM21, VARP and VPS29 cells. Cells were stained with an Alexa488 conjugated anti-mouse antibody. Arrowheads point to endosomal tubules. (C) Percentage of cells with tubules in parental or in various knockout cells; SEM, n=3 independent experiments. (D) Ratio of SLC3A2 integrated densities to GAPDH; SEM, n=3 independent experiments. (E-G) Retromer is required for endosomal localization of RAB21.

Immunofluorescence of transiently expressed GFP:RAB21 with endogenous (E) EEA1 or

(F) VPS26 in parental or VARP- and VPS29-deleted cells. Boxed region is magnified underneath. (G) Per cell Pearson Correlation between GFP:RAB21 and EEA1 or VPS26;

SEM, n=3 independent experiments.

Data information: (B) scale bars represent 20µm, (E and F) scale bars represent 10µm or

2.5µm in enlarged views. Statistical tests used: (C) – Unpaired t-tests, (D) One sample T- tests (G) - Mann-Whitney tests. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats.

Figure 8: Model of RAB21-mediated WASH/retromer cargo sorting. RAB21 is required for endosomal sorting of direct clathrin-independent cargos. RAB21 associates with

WASH/retromer subcomplexes at endosomes. RAB21 could either i) recruit

WASH/retromer or ii) be recruited by WASH/retromer or iii) be part of a positive feedback loop that would allow WASH/retromer and RAB21 recruitment at endosomes. Endosomal

RAB21 would be required for WASH-mediated F-actin polymerization. Although the data does not directly demonstrate a direct link between F-actin generation and cargo sorting, we propose that RAB21-dependent F-actin generation would be required for sorting of a CIE cargo subclass (MCT1, SLC3A2, Basigin and CD44), while it would not be required for other cargos (CI-MPR or Glut1). In RAB21 knockout cells, decreased WASH/retromer

117 118 endosomal localization is observed, which results in reduced endosomal F-actin and direct

CIE cargos misrouting. This ultimately leads to the lysosomal degradation of these misrouted cargos (demonstrated for SLC3A2). Dotted line between VARP and RAB21 indicates that their co-regulation with the retromer is speculative.

Expanded View Figure Legends

Figure EV1: Network of merged HCT116 and HeLa cells highlights new potential

RAB21 interactors with roles in trafficking. Anti-RAB21 immunoblot shows similar expression between GFP:RAB21 variants, which are over-expressed approximately 7-fold compared to endogenous RAB21 in (A) HeLa and (B) HCT116 cells. (C) Localization of

RAB21 variants in HeLa and HCT116 cells. Scale bars represent 10µm or 5µm in enlarged views. (D) Per cell Pearson Correlation between RAB21 variants and EEA1; Error bars are

SEM and Mann-Whitney tests were used for statistical comparisons, n=4 independent experiments. (E) VENN diagram highlighting the number of interactors identified with all

RAB21 variants in the two cell lines. (F) Cytoscape-generated interaction network; colors represent subcellular localization. (G) Reactome analysis of RAB21 interactors.

Figure EV2: AP-MS identification of RAB21 interactors and Gene Ontology enrichments of these interactors. VENN diagram highlighting the number of interactors identified with each RAB21 variant in (A) HeLa and (B) HCT116 cells. (C) Hierarchical clustering of all 29 core RAB21 interactors generated on Perseus. Reactome analysis of

RAB21 interactors in (D) HeLa and (E) HCT116. (F) Reactome enrichment established from the dataset published in [46].

118 119

Figure EV3: APEX2:RAB neighbors weakly overlap with a generic endosomal probe.

(A) Western blot analysis comparing APEX2:RAB expression levels and biotinylation efficiencies. Anti-myc immunoblot was performed to monitor RAB expression profiles. (B)

VENN diagram highlighting the number of common neighbors identified with each RAB21 variant. (C) ProHits-viz generated dot blots similar to Figure 2, but in which the

APEX2:2xFYVE dataset was included. Note that some proteins are missing compared to the

Figure 2 list given that they were not enriched in both 2xFYVE or RAB21 samples. Also, the relative abundance (size of the dots) on the dot plots is in accordance with all the plotted proteins and to the baits-depicted. Hence, dot size will be different from what is shown in

Figure 2C in order to facilitate the comparison of RAB21 to the 2xFYVE dataset. (D) Bait- bait comparison on ProHits-viz, between <0.01 FDR APEX2:RAB neighbors. The 60 preys with the highest relative abundance for each bait are represented. Red dots highlight named proteins, while weak blue dots represent proteins identified with a RAB not plotted in the graph where present. (E) Streptavidin Western blotting of total biotinylated proteins in

APEX2:RAB7a Flp-In/T-REx HeLa cells and APEX2:RAB7a biotinylated proteins

(streptavidin) are seen colocalizing with LAMP1 in HeLa cells. Scale bars represent 10µm or 5µm in enlarged view. n=2 independent experiments.

Figure EV4: Validation of specific APEX2:RAB21 hits and further characterization of

RAB21 knockout cells. ProHits-viz generated dotblots of RABs (A) phosphoinositide regulators and (B) potential new RAB21 interactors/neighbors. (C) SARA recovered from

APEX2:RAB21-mediated biotinylation. Immunoblot analysis of GFP:SARA and total biotinylated proteins. Lysates correspond to 5% input, n=3 independent experiments (D)

RAB21 interacts with PLEKHM2 and SLC7A11. FLAG immunoprecipitation of

119 120

FLAG:RAB21 and immunoblot of co-expressed PLEKHM2:HA and SLC7A11:HA. Lysates correspond to 5% input, n=3 independent experiments. (E) Schematic representation of the various indels or insertions observed in the gRNA-2 and gRNA-3 cell populations. PAM sequences are represented in red and the guide sequence in blue. The number of clones corresponds to the number of bacterial colonies sequenced that harbor the depicted modifications. (F) Autophagosomes accumulate in RAB21 knockout cells. Cell profiler quantification of the average number of LC3 puncta per cell; SEM, n=2 independent experiments. (G) Transferrin endocytosis is decreased in RAB21 KO cells. (H) Average transferrin-Alexa-647 intensity per cell following a 30 minutes chase; SEM. n=2 independent experiments.

Data Information: (G) Scale bars 100µm. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats.

Figure EV5: RAB21 regulates WASH and endosomal retromer recruitment, while

RAB21 knockout leads to increased SLC3A2 lysosome colocalization. (A-C) RAB21 knockout decreases SNX1 co-localization at endosomes without impairing vesicular SNX1 recruitment. (A) Immunofluorescence of endogenous SNX1 and EEA1 in RAB21-depleted cells. Boxed region is magnified and single channel images are depicted. (B) Per cell Pearson

Correlation between SNX1 and EEA1; SEM, n=2 independent experiments. (C) Average number of SNX1 puncta per cell; SEM, n=2 independent experiments. (D)

Immunofluorescence of endogenous FAM21 and EEA1 in RAB21 depleted cells. n=2 independent experiments. (E) Transient RAB21 wild type overexpression in RAB21 knockout cells rescues RAB21 deletion; SEM. n=3 independent experiments. Note that about fifty percent of cells were transfected in these rescue experiments, while all cells were

120 121 counted for tubules, explaining the lower number of tubules in the rescue compared to parental cells. (F) Colocalization of SLC3A2 with RAB7 and LAMP1 in RAB21 knockout cells. (G) Per cell Pearson Correlation between SLC3A2 and the two markers; SEM, n=2 independent experiments. (H) FACS analysis of SLC3A2 total protein levels in WT or

RAB21 knockout cells. n=3 independent experiments. (I) WB analysis of SLC3A2 total protein levels in WT or RAB21 knockout cells. Ratio of SLC3A2 integrated densities to

Tubulin; SEM. (J) C16orf62 endosomal localization does not depend on RAB21.

Immunofluorescence of endogenous c16orf62 and EEA1 in RAB21-depleted cells. n=3 independent experiments.

Data Information: Scale bars 10µm or 2.5µm in enlarged views. Statistical tests used: (E) -

Mann-Whitney tests, (I) – One sample T-tests. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats.

121 122

Article 1 Figure 1 : APEX2:RAB expression lead to endosomal biotinylation

122 123

Article 1 Figure 2 : Comparison of APEX2:RAB21-, 5-, 4- and 7-generated proteomes

123 124

Article 1 Figure 3 : RAB21 interacts and colocalizes with the WASH and retromer complexes.

124 125

Article 1 Figure 4 : Characterization of RAB21 knockout HeLa cells.

125 126

Article 1 Figure 5 : RAB21 modulates WASH and retromer endosomal recruitment and WASH activity.

126 127

Article 1 Figure 6 : RAB21 is required for endosomal sorting of specific cargo types.

127 128

Article 1 Figure 7 : WASH and retromer complexes are required for SLC3A2 and Basigin sorting, and for full RAB21 activation and endosomal localization.

128 129

Article 1 Figure 8 : Model of RAB21-mediated WASH/retromer cargo sorting.

129 130

Article 1 Figure EV 1 : Network of merged HCT116 and HeLa cells highlights new potential RAB21 interactors with roles in trafficking

130 131

Article 1 Figure EV 2 : AP-MS identification of RAB21 interactors and Gene Ontology enrichments of these interactors.

131 132

Article 1 Figure EV 3 : APEX2:RAB neighbors weakly overlap with a generic endosomal probe.

132 133

Article 1 Figure EV 4 : Validation of specific APEX2:RAB21 hits and further characterization of RAB21 knockout cells.

133 134

Article 1 Figure EV 5 : RAB21 regulates WASH and endosomal retromer recruitment, while RAB21 knockout leads to increased SLC3A2 lysosome colocalization.

134 135

3ARTICLE 2 RAB21 interacts with TMED10 and modulates its localization and abundance

Auteurs de l’article: Del Olmo T, Lacarrière-Keïta C, Normandin C, Jean D, Boisvert FM, Jean S.

Statut de l’article: publié dans Biology Open. 2019 Sep 9;8(9). pii: bio045336. doi: 10.1242

Avant-propos: Pour ce projet, j’ai généré les lignées HeLa-F/T et HCT116-F/T exprimant GFP:RAB21WT. J’ai réalisé les coimmunoprécipitations montrant une interaction entre TMED10 et GFP :RAB21 dans ces deux lignées. J’ai ensuite réalisé l’ensemble des expériences d’immunofluorescence montrant les différences de localisation de la protéine GFP:TMED10 entre les lignées sauvage et délétée pour RAB21. J’ai validé la baisse des niveaux d’expression de TMED10 dans les lignées KO-RAB21 par des expériences d’immunobuvardage. J’ai testé la stabilité de la protéine TMED10 par immunobuvardage dans les lignées KO-RAB21 traitées avec le cycloheximide. J’ai réalisé les expériences de purification GST ne montrant pas d’interaction directe entre RAB21 et TMED10. Enfin, j’ai participé à la rédaction du manuscrit et à la conception des figures. Ma contribution à cet article représente donc environ 90% du travail présenté.

Résumé : Le trafic membranaire contrôle le transport vésiculaire des cargos entre les différents compartiments cellulaires. Le trafic vésiculaire est essentiel pour l'homéostasie cellulaire et des défauts du trafic sont liés à plusieurs pathologies telles que les maladies neurodégénératives. Les endosomes précoces agissent comme des stations de triage des cargos internalisés qui acheminent ces cargos vers les divers organites cellulaires. Une classe importante de régulateurs du trafic membranaire sont les RAB GTPases. RAB21 a été associée à de multiples fonctions. En effet, RAB21 régule l'internalisation des intégrines, le tri endosomal de cargos indépendants de la clathrine et régule l'autophagie. Bien que RAB21 soit principalement associée aux endosomes précoces, il a été démontré qu'elle médie des événements de tri au Golgi. A partir de données de spectrométrie de masse, nous avons identifié une interaction entre RAB21 et TMED10 et 9, des régulateurs essentiels des vésicules de COPI et de COPII. A l'aide de cellules délétées pour RAB21, nous avons identifié le rôle de RAB21 dans la modulation de la localisation de TMED10 au Golgi. L'ensemble de notre étude suggère une fonction potentielle de RAB21 dans la modulation du trafic de TMED10, et donc un rôle pour RAB21 dans le développement de pathologies neurodégénératives.

135

136

RAB21 interacts with TMED10 and modulates its localization and

abundance

Tomas Del Olmo1, Camille Lacarrière-Keïta1, Caroline Normandin1, Dominique Jean1,

François-Michel Boisvert1 and Steve Jean1*

*Corresponding author:

Email: [email protected]

Telephone: 819-821-8000 Ext: 70450

FAX: 819-820-6831

1Faculté de Médecine et des Sciences de la Santé

Department of Anatomy and Cell Biology

Université de Sherbrooke

3201, Rue Jean Mignault

Sherbrooke, Québec, Canada, J1E 4K8

Running title: RAB21 binds TMED10

Keywords: RAB21, TMED10, TMED9, p24 family.

136 137

Summary statement A small early endosomal RAB GTPase is found to interact with p24 family members, with potential impacts on p24 functions.

137 138

Abstract

Membrane trafficking controls vesicular transport of cargo between cellular compartments.

Vesicular trafficking is essential for cellular homeostasis and dysfunctional trafficking is linked to several pathologies such as neurodegenerative diseases. Following endocytosis, early endosomes act as sorting stations of internalized materials, routing cargo toward various fates. One important class of membrane trafficking regulators are RAB GTPases.

RAB21 has been associated with multiple functions and regulates integrin internalization, endosomal sorting of specific clathrin-independent cargo and autophagy. Although RAB21 is mostly associated with early endosomes, it has been shown to mediate a specific sorting event at the Golgi. From mass spectrometry data, we identified a GTP-favored interaction between RAB21 and TMED10 and 9, essential regulators of COPI and COPII vesicles. Using

RAB21 knockout cells, we describe the role of RAB21 in modulating TMED10 Golgi localization. Taken together, our study suggests a new, potential function of RAB21 in modulating TMED10 trafficking, with relevance to neurodegenerative disorders.

138 139

Introduction

Membrane trafficking, which represents all vesicular exchanges between organelles and cellular compartments, is highly regulated and essential for cellular homeostasis (Vicinanza et al., 2008). Indeed, trafficking defects are involved in a large panel of diseases, such as neurological pathologies (Stenmark, 2009). One important class of membrane trafficking regulators are the RAB GTPases (Jean and Kiger, 2012). With almost 70 members, RABs represent the largest family of small GTPases in humans (Rojas et al., 2012). These proteins mediate each step of vesicular trafficking, from membrane budding to vesicle transport, to fusion with target organelles (Hutagalung and Novick, 2011). Given their roles in trafficking,

RABs are tightly regulated (Barr and Lambright, 2010). Thus, RABs cycle between their active GTP-bound form and inactive GDP-bound form. RABs are activated by GEFs

(guanine exchange factors), which catalyze the exchange of GDP to GTP and are inhibited by GAPs (GTPase activated proteins) that trigger intrinsic hydrolytic activity of RABs (Barr and Lambright, 2010). Once activated, RABs recruit a large number of effectors to achieve their functions (Grosshans et al., 2006). Moreover, RABs can directly interact with cargo to regulate their trafficking (Pellinen et al., 2006).

RAB21 regulates integrin internalization by binding directly to 51 integrin (Pellinen et al.,

2006). Initially described as an early endosomal RAB (Simpson et al., 2004), RAB21 has been shown to be involved in various specific functions. It mediates EGFR degradation

(Yang et al., 2012), controls neurite extensions (Burgo et al., 2009; Burgo et al., 2012), regulates autophagic flux (Jean et al., 2015) and was recently shown to be associated with clathrin-independent cargo trafficking via WASH and retromer complexes (Del Olmo et al.,

139 140

2019). In the same study, potential interactions between TMED10, TMED9 and RAB21 were observed by quantitative mass spectrometry analysis (Del Olmo et al., 2019).

TMED10 and TMED9 both belong to the p24 family of proteins (Pastor-Cantizano et al.,

2016). These proteins are mostly localized between the ER and Golgi compartments, cycling between both of them and mediating cargo transport through COPI and COPII vesicles

(Popoff et al., 2011). All proteins of the p24 family are composed of a GOLD intraluminal domain allowing interaction with cargo (Anantharaman and Aravind, 2002), a coiled-coil domain, a transmembrane domain mediating homo- and hetero-dimerization (Contreras et al., 2012) and a short cytosolic domain involved in protein sorting (Contreras et al., 2004).

The cytosolic domain has been demonstrated to interact with the GTP-bound ARF1 (Gommel et al., 1999), allowing recruitment of ERD2 and formation of COPI vesicles (Majoul et al.,

2001). Importantly, TMED10 and TMED2 expression is necessary to maintain ER and Golgi integrity (Montesinos et al., 2012). RAB21 has been shown to sort VAMP7 at the Golgi

(Burgo et al., 2012), and several members of the p24 family have been identified as potential

RAB21 binding proteins by mass spectrometry analysis (Del Olmo et al., 2019). Therefore, we characterized the TMED10 and RAB21 interaction using biochemical and genetic approaches. This allowed us to define a RAB21 requirement for appropriate TMED10 localization and protein abundance.

Results

TMED10 interacts indirectly with RAB21

Although the functional association of TMED10 with ARF1 has been well characterized

(D’Souza-Schorey and Chavrier, 2006), potential interactions with RAB family GTPases

140 141 have only been shown by proteomic analysis (Hein et al., 2015) or genetic screens (Blomen et al., 2015), but have not been further assessed. Our recent mass spectrometry data identified a potential interaction between RAB21, TMED10, and TMED9. Strong enrichment of these two proteins with wildtype or the GTP-bound form of RAB21 have been observed using quantitative interactomics experiments (Del Olmo et al., 2019), in both HeLa and HCT116 cells (Fig. S1A from (Del Olmo et al., 2019)).

To validate the interaction between RAB21 and TMED10, GFP co-immunoprecipitation assays were performed in Flp-In/T-REx HeLa and HCT116 cell lines that express

GFP:RAB21 close to endogenous levels in response to doxycycline treatment, as described previously (Del Olmo et al., 2019). Consistent with the proteomics data, endogenous

TMED10 was enriched in GFP:RAB21 immunoprecipitations in both, HeLa (Fig. 1A) and

HCT116 (Fig. 1B) cells, while an unrelated golgi protein, TGN38, was not (Fig. S1B). To test whether the identified interaction was direct, GST-pulldown assays were performed using purified GST:RAB21 or GST:RAB21-Q78L (GTP-bound) and incubated with HeLa cell lysates. Pulldown assays showed no specific enrichment of TMED10 or TMED2 with either wildtype or GTP-bound form, although VPS35 (Del Olmo et al., 2019) was present with GST-RAB21 (Fig. S1C and D). From these results, we conclude that RAB21 interacts indirectly with TMED10.

TMED10 interacts preferentially with activated RAB21

To assess the RAB21 interaction with TMED10, we performed proximity ligation assays

(PLA). HeLa cells were singly or co-transfected with TMED10:3xHA and either V5:RAB21-

WT, V5:RAB21-Q78L (GTP-bound) or V5:RAB21-T33N (GDP-bound) variants. The

141 142 number of PLA puncta per cell (indicative of TMED10 and RAB21 proximity) was counted through confocal imaging and automated image analysis. While transfection of either

TMED10 or RAB21 alone yielded a maximum of 9 puncta per cell (Fig. 2A and B), co- transfection of TMED10 with RAB21-WT or RAB21-Q78L led to a considerable increase in the number of PLA puncta per cell, reaching an average of 42 puncta per cell in RAB21-

WT cells. Notably, the number of PLA puncta per cell was significantly higher in RAB21-

WT and RAB21-Q78L variants compared to RAB21-T33N (Fig. 2A and C). These PLA results are in accordance with the proteomics data and indicate that the interaction between

RAB21 and TMED10 is increased upon RAB21 activation.

RAB21 knockout affects TMED10 localization in cells

TMED10 has been reported to localize at the ER-Golgi interface. On the other hand, RAB21 localizes mostly on early endosomes except for the dominant negative RAB21 (T33N), which is strongly associated with the Golgi (Simpson et al., 2004). A previous study identified VARP- and RAB21-dependent functions in VAMP7 trafficking at the Golgi in neuronal cells (Burgo et al., 2012). Therefore, we assessed the functional relationship between RAB21 and TMED10 in RAB21 knockout cells. Phenotype specificity was ensured by generating two independent cell populations using two independent guide RNAs. These cells have previously been validated by sequencing and western blot analyses (Del Olmo et al., 2019).

Parental and RAB21 knockout cells were transfected with GFP:TMED10 (Blum and Lepier,

2008) and TMED10 localization at cis- and trans-Golgi was investigated by colocalization with GM130, TGN46, and ci-MPR, respectively (Fig. 3A, S2A and S2B). Interestingly,

142 143

RAB21 deletion reduced TMED10 localization in the cis-Golgi compartment (Fig. 3A).

Multiple TMED10 puncta were observed outside the cis-Golgi in these cells. Moreover,

TMED10 colocalization with GM130 significantly decreased in both HeLa-RAB21 KO cells compared to parental cells (Fig. 3B). TMED10 colocalization with TGN46 also decreased in both HeLa-RAB21-KO cell populations, however only the gRNA-2 population showed a statistical difference (Fig. S2C). Surprisingly, no difference was observed between TMED10 and ci-MPR colocalization (Fig. S2D). Given that ci-MPR also labels endocytic vesicles, an increased localization of TMED10:GFP in these vesicles could potentially compensate for the observed difference at the Golgi and yield similar Pearson correlation values. Taken together, these results suggest the role of RAB21 in TMED10 maintenance or targeting at the cis-Golgi compartment and potentially at the trans-Golgi as well.

RAB21 depletion reduces TMED10 protein levels

Given that TMED10 was mis-localized in RAB21 knockout cells, we assessed whether

TMED10 protein levels were also affected. Using western blotting, we compared relative

TMED10 expression in parental and RAB21-KO HeLa cells (Fig. 4A). Relative protein quantification showed that in both RAB21 knockout populations, TMED10 expression was almost twice as low as in the control (Fig. 4B). Since p24 family members are known to oligomerize (Contreras et al., 2012) and depletion of TMED10 affects other p24 family members (Pastor-Cantizano et al., 2016), we also analyzed TMED2 protein abundance. In accordance with the result observed for TMED10, we found that TMED2 expression was also altered in RAB21 knockout cells (Fig. 4C). We assessed if this was due to changes in transcription or in protein stability. Quantitative PCR analyses of TMED2 and TMED10 did not highlight any significant difference in expression (Fig. 4D and E). Similarly, a

143 144 cycloheximide chase, did not show apparent differences in stability (Fig. 4F and G). We further assessed if TMED10 half-life was modulated by proteasome or lysosomal degradation. MG132 or Bafilomycin A1 treatments, which block proteasome or lysosome functions respectively, did not significantly impact TMED10 protein levels in either parental or RAB21 knockout cells (Fig. S3A and B). From these results, we conclude that RAB21 is required for expression of TMED10 and TMED2, that TMED10 has a long half-life in HeLa cells, and that the exact mechanism leading to TMED2 and 10 downregulation needs to be elucidated.

Discussion

Firstly identified to play a specific role in integrin trafficking (Pellinen et al., 2006), RAB21 is now associated with several other functions (Del Olmo et al. 2019; Jean et al. 2015; Alanko et al. 2015). In the present study, we confirmed previous mass spectrometry data showing a potential interaction between RAB21 and TMED10 (Del Olmo et al., 2019). We conclude that activated RAB21 interacts indirectly with TMED10 as evidenced by immunoprecipitation analysis, proximity ligation assays and pull-down data. We further show that RAB21 is required for TMED10 localization in the Golgi and that RAB21 influences TMED10 and TMED2 protein expression.

TMED10 is mostly observed at the cis-Golgi (Pastor-Cantizano et al., 2016), compared to

RAB21, which is mostly endosomal. However, RAB21 has been observed at the Golgi in neurons where it regulates, through VARP, the sorting of VAMP7 (Burgo et al., 2012).

Hence, it is possible that another function of VARP would be regulation of RAB21 interaction with TMED10. The preferential interaction of TMED10 with activated RAB21

144 145 thus suggests that TMED10 could act as a RAB21 effector or that activated RAB21 could influence the interaction between TMED10 and its specific cargo. The data from RAB21 knockout cells indicates that both possibilities are plausible. However, it is unlikely that

TMED10 would act as a RAB21 effector, due to the lack of a direct interaction between the two proteins. Hence, a Golgi-associated RAB21 pool could retain TMED10 at the Golgi, by interacting with TMED10, possibly through an unknown protein. Alternately, RAB21 could strengthen or weaken TMED10 interaction with cargo, as observed for ARF1 (Luo et al.,

2007) and RAB10 (Wang et al., 2010).

TMED10 has been observed to be localized at other intracellular compartments such as the

ER, the ERGIC compartment, on secretory vesicles and at the plasma membrane (Pastor-

Cantizano et al., 2016). Given the known functions of RAB21 in endocytosis and protein sorting (Jean et al. 2015; Pellinen et al. 2006; Del Olmo et al. 2019), and considering the observed RAB21-dependent TMED10 localization at the Golgi, it is also possible that

RAB21 could be involved in TMED10 recycling from the plasma membrane to the Golgi. A recent study suggests that TMED10 cycles through the plasma membrane with improperly folded GPI-anchored proteins (Zavodszky and Hegde, 2019). Hence, an interesting possibility would be that RAB21 is involved in regulating the trafficking and degradation of improperly folded GPI-anchored proteins, and as such RAB21 depletion would lead to improper TMED10 localization.

In our recent proteomic study where we noticed RAB21/TMED10 interaction, we did not observe any significant enrichment of TMED10 in APEX2:RAB21-mediated proximity labeling. This was rather surprising given the large number of proximal RAB21 proteins

145 146 identified in that study (Del Olmo et al., 2019). We believe that this could be explained by the fact that APEX2 biotinylation occurs mostly on tyrosine residues (Lee et al., 2017), and p24 family members contain only short cytosolic domains with no tyrosine (Pastor-Cantizano et al., 2016). Therefore, although TMED10 could still be proximal to APEX2:RAB21 in these experiments, it might not be biotinylated and hence detected. This indicates the necessity to combine experimental approaches to define protein interactomes. In this regard, both SILAC and proximity labeling approaches will complement each other in future studies.

RAB21 modulates TMED10 and TMED2 protein levels. TMED10 was found to have a half- life of 3 hours in the neurons (Liu et al., 2008), while TMED2 was shown to have a very long half-life in Vero cells (Füllekrug et al., 1999). TMED10 degradation in neurons was mostly through the proteasome, while the mechanism in Vero cells has not been characterized.

Hence, it remains unclear whether a common pathway is responsible for the degradation of p24 family members. From our data, we could not identify the mechanism which contributes to the decreased abundance of TMED10 and TMED2. We observed that TMED10 had a long half-life in HeLa (>7 h) and we could not detect a shorter half-life in RAB21 knockout cells.

Furthermore, proteasome or lysosome inhibition for 16 hours did not strongly affect

TMED10 levels. We did observe a slight increase in TMED10 levels upon lysosome inhibition (Fig. S3B), but this observation was not consistent over the various repeats. Hence, decreased protein levels of TMED10 and 2 could be attributed to an indirect effect on other p24 family members or to a slight increase in degradation kinetics that could not be observed in the timescale of our experiment. We ruled out a general effect on global protein translation, since multiple proteins are not downregulated in RAB21 knockout cells (Del Olmo et al.,

2019).

146 147

TMED10 plays a dual role in cargo trafficking. By itself, TMED10 is involved in specific secretion of GPI-anchored proteins (Theiler et al., 2014) or PAR-2 (Zhao et al., 2014) towards the plasma membrane. On the other hand, TMED10 has been shown to retain cargo such as MHC class I (Jun and Ahn, 2011) in the ER, and PKC- in Golgi-like structures

(Wang et al., 2011), thus inhibiting their activities. Similarly, TMED10 is responsible for direct -secretase inhibition (Pardossi-Piquard et al., 2009). We hypothesized that RAB21 could modulate general TMED10-regulated cargo trafficking by disturbing TMED10 localization and stability. With respect to this hypothesis, RAB21 has been shown to interact with Presenilin 1, reducing APP synthesis through -secretase inhibition (Sun et al., 2017).

Hence, it would be interesting to investigate if the involvement of RAB21 and TMED10 in

-secretase inhibition are linked or independent. Moreover, inhibition of TMED10 expression is involved in ATG4-mediated autophagy activation (Shin et al., 2019). Hence, in addition to its published role in modulating autophagosome-lysosome fusion through

VAMP8 trafficking (Jean et al., 2015), RAB21 could also influence other steps of autophagy via its interaction with TMED10.

In this study, we validated the interaction between RAB21 and TMED10, a cargo adaptor, and showed the role of RAB21 in modulating TMED10 localization. Considering that

RAB21 and TMED10 are involved in the regulation of autophagy and -secretase activity, respectively, and the fact that both of these pathways are associated with Alzheimer’s disease, it would be interesting to further investigate the functional relationship between these two trafficking proteins in Alzheimer’s disease. We believe that this study provides a framework

147 148 for further studies on the association between the early endosomal RAB GTPase RAB21 and p24 family members.

Materials and Methods

Cell culture

All cell lines were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with penicillin, streptomycin and 10% fetal bovine serum (Wisent) under 5% CO2 at 37°C.

HeLa and HCT116 Flip-In/T-rex cells and RAB21 knockout cell populations were described previously in (Del Olmo et al., 2019).

Generation of DNA constructs

PCDNA3-TMED10:3xHA was generated by amplifying TMED10 by PCR from HeLa cDNA generated using the Superscript III First-Strand synthesis kit. The TMED10 PCR fragment was ligated into pCDNA3-3xHA using the In-Fusion cloning kit (Clontech). The

GFP-TMED10 plasmid was a kind gift of Robert Blum (Blum and Lepier, 2008). All constructs were validated by sequencing.

Immunoprecipitations

3.5x106 HCT116 or HeLa cells were plated in 100 mm dishes and grown for 24 hours with

11 ng/ml of doxycycline to allow GFP:RAB21 induction. HCT116 cells were washed twice with cold PBS and lysed on ice for 20 min with 1 mL CoIP Buffer (1% IGEPAL CA-630, 1 mM EDTA, 150 mM NaCl, 0.1 mM EGTA, 25 mM Tris–HCl pH 7.4, 15 mM MgCl2, 2 mM

Na3VO4, 10% glycerol, 2× protease inhibitors) per plate. HeLa cells were fixed 15 min at room temperature with 0.5 % formaldehyde with gentle rocking, following which

148 149 formaldehyde was quenched for 5 min with 125 mM glycine at room temperature. Fixed

HeLa cells were washed twice with cold PBS and lysed for 15 min at 4°C on a rotator with

1 mL of lysis buffer (1% IGEPAL CA-630, 1mM EDTA, 150 mM NaCl, 25 mM Tris pH

7.4, 5% glycerol). For both, HeLa and HCT116 cells, remaining membrane aggregates and

DNA were removed by centrifugation at 16,000 g for 12 min at 4°C. Protein concentrations were determined using a BSA assay (Pierce), and immunoprecipitations were performed in individual tubes with equivalent quantities of proteins. 15l of GFP-Trap beads

(ChromoTek) were used for individual IPs. For HCT116 cells, immunoprecipitations were performed on a rotator for 2.5 hours at 4°C. Beads were then washed twice with CoIP buffer and twice again with CoIP buffer lacking IGEPAL CA-630. Immunoprecipitations in HeLa cells were carried out for 4 hours on a rotator at 4°C. Following this incubation, beads were washed 3 times with lysis buffer. Finally, for both HCT116 and HeLa cells, excessive wash buffer was removed from the beads at the end of the immunoprecipitation protocol and 25

L of 2× SDS loading buffer was added to each sample to elute proteins from beads.

Pulldown assays

3.5x106 HeLa cells were grown for 24 hours in 100 mm plates after which cells were lysed for 20 min on ice using 1 mL of MLB modified buffer (1% IGEPAL CA-630, 10% glycerol,

100 M EGTA and 100 M GTP, 25 mM HEPES, 150 mM NaCl, 20 mM MgCl2, and 1 mM sodium orthovanadate) supplemented with 2x protease inhibitors. Lysates were cleared by centrifugation at 16,000 g for 12 min at 4°C. GST and GST:RAB21 were purified following

(Jean et al., 2012). Prior to the pulldowns, GST and GST:RAB21 beads were washed three times in MLB modified buffer minus IGEPAL CA-630 and incubated on a rotator at 4°C for

149 150

20 min in MLB modified buffer lacking IGEPAL CA-630. Beads were further washed three times in complete MLB modified buffer and 900 l of HeLa cell lysates was added to the beads for each pulldown, this was incubated for 1 hour at 4°C on a rotator. Following this incubation, beads were washed three times with MLB modified buffer containing 0.2%

IGEPAL CA-630. Protein were eluted with 30 L of 2× SDS loading buffer.

Immunoblots

For immunoblot analyses, 3x105 parental HeLa cells and 4.5x105 HeLa RAB21 KO cells were grown for 24 hours in 6-well plates. Cells were lysed with 200 L of CoIP buffer as described above for immunoprecipitations. Lysates were quantified and the same amounts of proteins were used for analysis. Proteins were separated on 4–20% TGX precast gels (Bio-

Rad) and transferred onto PVDF membranes (Millipore) using the trans-blot turbo system from Bio-Rad. Antibodies used for immunoblotting were anti-RAB21 (1:1,000, Invitrogen

#PA5-34404), anti-GFP (1:500, Santa Cruz #9996), anti-TMED10 (1:1000, Abcam

#134948), anti-TMED2 (1:1000, Santa Cruz #376458), anti-GAPDH (1:8000, Cell signaling

#8884), anti-LC3B (1:1000, Cell Signaling #3868), anti-TGN38 (1:1000, Santa Cruz

#166594), anti-Ubiquitin (1:1000, Cell Signaling 3933), anti-Vps35 (1:500, Santa Cruz

#374372) and anti-rabbit and mouse HRP (1:10,000, Jackson Laboratories #115-035-144 and

#115-035-146, respectively). Membranes were imaged on a Bio-Rad Chemidoc XR station following 5 min incubation with Luminata Forte (Millipore) or Clarity Max chemiluminescent substrates (Bio-Rad). On specific occasions, membranes were cut to allow probing with multiple antibodies simultaneously.

150 151

TMED10 stability experiments were performed by incubating cells in full media containing either 25µg/ml cycloheximide or 10 µM MG132 for the indicated amount of time. For

Bafilomycin A1 treatments, 0.2 µg/ml and 0.1µg/ml were used for the 4h and the 16h time point respectively. A lower concentration was used for the 16h time point due to BafA1 toxicity. Cells were lysed and proteins immunoblotted as described above.

Immunofluorescence, colocalization and proximity ligation assay

A total of 20,000 wild type HeLa or 50 000 RAB21 knockout cells were plated on glass coverslips (#1.5) in 24-wells plate and cultured overnight. The following day, pcDNA3-

GFP:TMED10 or pcDNA3-TMED10:3xHA with or without pcDNA3-V5:RAB21 were transfected using Jetprime (Polyplus) following manufacturer’s instructions. 24 hours following transfection, cells were washed twice with 1×PBS and fixed for 15 min at room temperature with 250 L of 4% paraformaldehyde in PBS. Cells were then washed three times 5 min each wash with 1x PBS. Fixed cells were blocked, and permeabilized for 60 min with 300 L of 5% goat serum and 0.3% Triton X-100 in PBS. Cells were then incubated overnight in a humidified chamber at 4°C with primary antibody in 1x PBS containing 0.3%

Triton X-100 and 1% BSA. The following day, primary antibodies were washed three times

5 min with 1x PBS. For immunofluorescences, cells were incubated 1 hour in a humidified chamber with secondary antibody at room temperature in the same buffer as the primary antibody. Secondary antibodies were washed three times 5-min with 1x PBS at room temperature. and cells were mounted in DAPI-containing mounting media (Sigma). For PLA, after having washed the primary antibodies, cells were incubated in a humidified chamber for 1 hour at 37°C with Sigma probes (+) and (-) in the 1X sigma dilution solution. Sigma

151 152 probes were washed twice 5 min with 1x Wash Buffer A. The ligation step was performed at

37°C for 30 min followed by two washes with buffer A. Amplification was performed for

100 min at 37°C in the dark. Cells were washed at room temperature twice for 10 min each wash with 1X wash buffer B and finally for 1 min in 0.01x wash buffer B and mounted in

DAPI-containing mounting media (Sigma). Antibodies used for immunofluorescences were anti-GM130 (1:100, Cell Signaling #12480), anti-ciMPR (1:100, Bio-Rad #MCA2048T) and anti-TGN46 (1:100, Novus Biological #NBP1-49643SS) and for PLA were anti-HA (1:1000,

Cell Signaling #3724) and anti-V5 (1:5000, Sigma-Aldrich #V8012).

Image analysis and statistics

All images were acquired on an Olympus FV1000 using a 63x 1.42NA plan Apo N objective or on a Zeiss LSM880 using a 40x 1.4NA plan Apo objective. Imaging settings were selected to minimize pixel saturation and to ensure proper Pearson correlation calculation. For figure preparation, all microscopy images were tresholded and cropped on Photoshop and assembled using Illustrator. All images were treated similarly, and only linear modifications were performed. The number of PLA puncta per cell was established using Cell Profiler.

Briefly, a pipeline allowing transfected cell identification (using GFP signal from cotransfection of a small amount of pEGFP-C1 together with the other plasmids) was used and the number of PLA puncta per EGFP positive cell was established using the relate and filter modules of Cell Profiler. Pearson correlations were also generated using Cell Profiler, with the distinction that cells were manually identified. Immunoblot data densitometry and image processing were performed using Image Lab (Bio-Rad). Immunoblot images were cropped and assembled using Photoshop and Illustrator respectively. For statistical analyses of images or immunoblots, every samples were first subjected to a normality test (if a

152 153 sufficient number of values were present), following which either unpaired t-test, Mann-

Whitney tests or One-sample t-tests were performed.

Quantitative PCR analysis

Wild type or RAB21 knockout cells were grown to 80% confluency in full media. CDNAs were prepared using the Maxima First Strand cDNA synthesis kit for RT-qPCR with dsDNAse (Thermo Fisher) following manufacturer’s instruction. Luna Universal qPCR

Master Mix was used for amplification and the reactions were performed on a Roche

LightCycler 96. Relative mRNA levels were calculated using the ∆∆Ct method and normalized to GAPDH. TMED10, 2 and GAPDH primers were predesigned qPCR primers obtained from IDT.

Acknowledgments

We are grateful to Robert Blum (Universitätsklinikum Würzburg) for kindly providing the

TMED10:GFP (Blum and Lepier, 2008) construct used throughout this study. We thank

Marie-Josée Boucher and Benoit Marchand for suggestions with the MG132 assay. We thank the proteomic and photonic microscopy core at the Université de Sherbrooke for proteomic services and confocal use. Steve Jean and Francois-Michel Boisvert are members of the

FRQS-Funded Centre de Recherche du CHUS. Steve Jean is a recipient of a Research Chair from the Centre de recherche médicale de l’Université de Sherbrooke (CRMUS).

Competing interests

No competing interests declared

153 154

Funding

This research was supported by operating grants from the Cancer Research Society (CRS –

22247) and the Canadian Institutes of Health Research (CIHR – 142305) and by junior faculty salary awards from Canadian Institutes of Health Research (CIHR) and Fonds de

Recherche du Québec-Santé (FRQS) to S.J.

References

Alanko, J., Mai, A., Jacquemet, G., Schauer, K., Kaukonen, R., Saari, M., Goud, B. and Ivaska, J. (2015). Integrin endosomal signalling suppresses anoikis. Nat. Cell Biol. 17, 1412–1421. Anantharaman, V. and Aravind, L. (2002). The GOLD domain, a novel protein module involved in Golgi function and secretion. Genome Biol. 3, research0023. Barr, F. and Lambright, D. G. (2010). Rab GEFs and GAPs. Curr. Opin. Cell Biol. 22, 1– 10. Blomen, V. A., Májek, P., Jae, L. T., Bigenzahn, J. W., Nieuwenhuis, J., Staring, J., Sacco, R., van Diemen, F. R., Olk, N., Stukalov, A., et al. (2015). Gene essentiality and synthetic lethality in haploid human cells. Science 350, 1092–6. Blum, R. and Lepier, A. (2008). The Luminal Domain of p23 (Tmp21) Plays a Critical Role in p23 Cell Surface Trafficking. Traffic 9, 1530–1550. Burgo, A., Sotirakis, E., Simmler, M. C., Verraes, A., Chamot, C., Simpson, J. C., Lanzetti, L., Proux-Gillardeaux, V. and Galli, T. (2009). Role of Varp, a Rab21 exchange factor and TI-VAMP/VAMP7 partner, in neurite growth. EMBO Rep. 10, 1117–1124. Burgo, A., Proux-Gillardeaux, V., Sotirakis, E., Bun, P., Casano, A., Verraes, A., Liem, R. K. H., Formstecher, E., Coppey-Moisan, M. and Galli, T. (2012). A Molecular Network for the Transport of the TI-VAMP/VAMP7 Vesicles from Cell Center to Periphery. Dev. Cell 23, 166–180. Contreras, I., Ortiz-Zapater, E. and Aniento, F. (2004). Sorting signals in the cytosolic tail of membrane proteins involved in the interaction with plant ARF1 and coatomer.

154 155

Plant J. 38, 685–98. Contreras, F.-X., Ernst, A. M., Haberkant, P., Björkholm, P., Lindahl, E., Gönen, B., Tischer, C., Elofsson, A., von Heijne, G., Thiele, C., et al. (2012). Molecular recognition of a single sphingolipid species by a protein’s transmembrane domain. Nature 481, 525–9. D’Souza-Schorey, C. and Chavrier, P. (2006). ARF proteins: Roles in membrane traffic and beyond. Nat. Rev. Mol. Cell Biol. 7, 347–358. Del Olmo, T., Lauzier, A., Normandin, C., Larcher, R., Lecours, M., Jean, D., Lessard, L., Steinberg, F., Boisvert, F. and Jean, S. (2019). APEX2‐mediated RAB proximity labeling identifies a role for RAB21 in clathrin‐independent cargo sorting. EMBO Rep. 20, e47192. Füllekrug, J., Suganuma, T., Tang, B. L., Hong, W., Storrie, B. and Nilsson, T. (1999). Localization and recycling of gp27 (hp24gamma3): complex formation with other p24 family members. Mol. Biol. Cell 10, 1939–55. Gommel, D., Orci, L., Emig, E. M., Hannah, M. J., Ravazzola, M., Nickel, W., Helms, J. B., Wieland, F. T. and Sohn, K. (1999). p24 and p23, the major transmembrane proteins of COPI-coated transport vesicles, form hetero-oligomeric complexes and cycle between the organelles of the early secretory pathway. FEBS Lett. 447, 179–185. Grosshans, B. L., Ortiz, D. and Novick, P. (2006). Rabs and their effectors: achieving specificity in membrane traffic. Proc. Natl. Acad. Sci. U. S. A. 103, 11821–11827. Hein, M. Y., Hubner, N. C., Poser, I., Cox, J., Nagaraj, N., Toyoda, Y., Gak, I. A., Weisswange, I., Mansfeld, J., Buchholz, F., et al. (2015). A Human Interactome in Three Quantitative Dimensions Organized by Stoichiometries and Abundances. Cell 163, 712–723. Hutagalung, A. H. and Novick, P. J. (2011). Role of Rab GTPases in membrane traffic and cell physiology. Physiol. Rev. 91, 119–149. Jean, S. and Kiger, A. A. (2012). Coordination between RAB GTPase and phosphoinositide regulation and functions. Nat. Rev. Mol. cell Biol. 13, 463–470. Jean, S., Cox, S., Schmidt, E. J., Robinson, F. L. and Kiger, A. (2012). Sbf/MTMR13 coordinates PI(3)P and Rab21 regulation in endocytic control of cellular remodeling. Mol. Biol. Cell 23, 2723–2740.

155 156

Jean, S., Cox, S., Nassari, S. and Kiger, A. A. (2015). Starvation-induced MTMR13 and RAB21 activity regulates VAMP8 to promote autophagosome-lysosome fusion. EMBO Rep. 16, 297–311. Jun, Y.-S. and Ahn, K.-S. (2011). Tmp21, a novel MHC-I interacting protein, preferentially

binds to β 2 -microglobulin-free MHC-I heavy chains. BMB Rep. 44, 369–374. Lee, S.-Y., Kang, M.-G., Shin, S., Kwak, C., Kwon, T., Seo, J. K., Kim, J.-S. and Rhee, H.-W. (2017). Architecture Mapping of the Inner Mitochondrial Membrane Proteome by Chemical Tools in Live Cells. J. Am. Chem. Soc. 139, 3651–3662. Liu, S., Bromley-Brits, K., Xia, K., Mittelholtz, J., Wang, R. and Song, W. (2008). TMP21 degradation is mediated by the ubiquitin-proteasome pathway. Eur. J. Neurosci. 28, 1980–1988. Luo, W., Wang, Y. and Reiser, G. (2007). p24A, a type I transmembrane protein, controls ARF1-dependent resensitization of protease-activated receptor-2 by influence on receptor trafficking. J. Biol. Chem. 282, 30246–30255. Majoul, I., Straub, M., Hell, S. W., Duden, R. and Söling, H. D. (2001). KDEL-cargo regulates interactions between proteins involved in COPI vesicle traffic: measurements in living cells using FRET. Dev. Cell 1, 139–153. Montesinos, J. C., Sturm, S., Langhans, M., Hillmer, S., Marcote, M. J., Robinson, D. G. and Aniento, F. (2012). Coupled transport of Arabidopsis p24 proteins at the ER- Golgi interface. J. Exp. Bot. 63, 4243–61. Pardossi-Piquard, R., Böhm, C., Chen, F., Kanemoto, S., Checler, F., Schmitt-Ulms, G., St George-Hyslop, P., Fraser, P. E., St. George-Hyslop, P. and Fraser, P. E. (2009). TMP21 transmembrane domain regulates γ-secretase cleavage. J. Biol. Chem. 284, 28634–28641. Pastor-Cantizano, N., Montesinos, J. C., Bernat-Silvestre, C., Marcote, M. J. and Aniento, F. (2016). p24 family proteins: key players in the regulation of trafficking along the secretory pathway. Protoplasma 1–19. Pellinen, T., Arjonen, A., Vuoriluoto, K., Kallio, K., Fransen, J. A. M. M. and Ivaska, J. (2006). Small GTPase Rab21 regulates cell adhesion and controls endosomal traffic of beta1-integrins. J. Cell Biol. 173, 767–780. Popoff, V., Adolf, F., Brügger, B. and Wieland, F. (2011). COPI budding within the Golgi

156 157

stack. Cold Spring Harb. Perspect. Biol. 3, a005231. Rojas, A. M., Fuentes, G., Rausell, A. and Valencia, A. (2012). Evolution: The Ras protein superfamily: Evolutionary tree and role of conserved amino acids. J. Cell Biol. 196, 189–201. Shin, J. H., Park, S. J., Jo, D. S., Park, N. Y., Kim, J. B., Bae, J.-E., Jo, Y. K., Hwang, J. J., Lee, J.-A., Jo, D.-G., et al. (2019). Down-regulated TMED10 in Alzheimer disease induces autophagy via ATG4B activation. Autophagy 1–11. Simpson, J. C., Griffiths, G., Wessling-Resnick, M., Fransen, J. A. M., Bennett, H. and Jones, A. T. (2004). A role for the small GTPase Rab21 in the early endocytic pathway. J. Cell Sci. 117, 6297–6311. Stenmark, H. (2009). Rab GTPases as coordinators of vesicle traffic. Nat. Rev. Mol. cell Biol. 10, 513–525. Sun, Z., Xie, Y., Chen, Y., Yang, Q., Quan, Z., Dai, R. and Qing, H. (2017). Rab21, a Novel PS1 Interactor, Regulates γ-Secretase Activity via PS1 Subcellular Distribution. Mol. Neurobiol. Theiler, R., Fujita, M., Nagae, M., Yamaguchi, Y., Maeda, Y. and Kinoshita, T. (2014). The α-helical region in p24γ2 subunit of p24 protein cargo receptor is pivotal for the recognition and transport of glycosylphosphatidylinositol-anchored proteins. J. Biol. Chem. 289, 16835–43. Vicinanza, M., D’Angelo, G., Di Campli, A. and De Matteis, M. A. (2008). Function and dysfunction of the PI system in membrane trafficking. EMBO J. 27, 2457–2470. Wang, D., Lou, J., Ouyang, C., Chen, W., Liu, Y., Liu, X., Cao, X., Wang, J. and Lu, L. (2010). Ras-related protein Rab10 facilitates TLR4 signaling by promoting replenishment of TLR4 onto the plasma membrane. Proc. Natl. Acad. Sci. 107, 13806– 13811. Wang, H., Xiao, L. and Kazanietz, M. G. (2011). p23/Tmp21 Associates with Protein Kinase Cδ (PKCδ) and Modulates Its Apoptotic Function. J. Biol. Chem. 286, 15821– 15831. Yang, X., Zhang, Y., Li, S., Liu, C., Jin, Z., Wang, Y., Ren, F. and Chang, Z. (2012). Rab21 attenuates EGF-mediated MAPK signaling through enhancing EGFR internalization and degradation. Biochem. Biophys. Res. Commun. 421, 651–657.

157 158

Zavodszky, E. and Hegde, R. S. (2019). Misfolded GPI-anchored proteins are escorted through the secretory pathway by ER-derived factors. Elife 8,. Zhao, P., Metcalf, M. and Bunnett, N. W. (2014). Biased Signaling of Protease-Activated Receptors. Front. Endocrinol. (Lausanne). 5, 67.

158 159

Figure legends

Figure 1. RAB21 interacts with TMED10

A) and B) Western blotting showing GFP-trap immunoprecipitation in HeLa (A) and

HCT116 (B) cells. Endogenous TMED10 was blotted and showed a specific enrichment with

GFP:RAB21 compared to FT control in both cell lines. Lysates represents 2% of input and n=4 independent experiments performed in each cell line.

Figure 2. RAB21 nucleotide status modulates the interaction with TMED10

A) PLA immunofluorescence showing respective proximities between TMED10:HA and either V5:RAB21 WT, Q78L or T33N variants. TMED10:HA and V5:RAB21 only are controls. PLA puncta are stained in red and nucleus in blue. Dotted lines define individual cells. Scale bar 10 m, n=2 independent experiments. B) Quantification of PLA controls showed in (A). Histograms represents average number of PLA puncta per cell, error bars are

SEM. No statistical analysis was performed. (C) Quantification of PLA experiments shown in (A). Histograms represent average number of PLA dots per cell in each RAB21 variant conditions, error bars are SEM. Mann-Whitney tests were used for statistical analysis.

Figure 3. TMED10 Golgi localization is altered in RAB21 knockout cells

(A) TMED10:GFP colocalization with endogenous GM130, in parental and RAB21 knockout cells. gRNA-2 and -3 are two independent populations of RAB21 knockout HeLa cells. TMED10:GFP is stained in green, GM130 in red and nucleus in blue. Scale bar 10 m.

Dotted squares are magnified, scale bar 5 m, n=3 independent experiments. (B)

Quantification of TMED10 and GM130 colocalization. Histogram represents average

159 160

Pearson correlation per cell, error bars are SEM. Mann-Whitney tests were used for statistical analysis.

Figure 4. RAB21 knockout reduces TMED10 and TMED2 protein levels

(A) Western blotting showing TMED10 and TMED2 protein levels in parental and RAB21 knockout cells. Endogenous TMED10, TMED2, RAB21 and Tubulin were blotted. Tubulin was used as housekeeping gene, n=4 independent experiments. (B) and (C) Quantification of relative protein expression showed in (A). TMED10/Tubulin (B) and TMED2/Tubulin (C) protein ratios were normalized to parental cells, error bars are SEM. One-sample t-tests were used for statistical analysis. (D and E) Quantitative PCR analysis of TMED10 and TMED2 transcripts respectively, n=4 independent experiments, error bars are SEM. Unpaired t-tests were used for statistical analysis. No statistical differences were observed and are thus not displayed. (F and G) TMED10 stability assay. Cycloheximide chases were performed to monitor TMED10 stability in parental and RAB21 gRNA-3 KO HeLa cells, n=3 independent experiments. (F) Endogenous TMED10, RAB21 and GAPDH were assessed through western blotting. (G) Quantification of relative protein expression showed in (F), error bars are SEM.

Unpaired t-tests were performed and no statistical differences were observed and therefore are not displayed.

Figure S1. Indirect interaction between RAB21 and TMED10

(A) SILAC enrichment ratios of GFP:RAB21 variants over the FT control in HeLa and

HCT116 cells, data from (Del Olmo et al., 2019). (B) Western blotting showing GFP-trap immunoprecipitation in HeLa cells. Endogenous TMED10 and TGN38 were blotted. A specific enrichment with GFP:RAB21 compared to FT control was only observed with

160 161

TMED10. Lysates represents 2% of input and n=3 independent experiments. (C) Western blotting showing GST:RAB21, GST:RAB21-Q78L pull-downs and GST only as control.

TMED10 and TMED2 were blotted and showed no specific enrichment with neither RAB21 or RAB21-Q78L compared to control. Gels were stained with coomasie blue to compare purified protein levels in all conditions, n=2 independent experiments. (D) Western blotting showing GST:RAB21 loaded with GTP, pull-downs and GST only as control. TMED10 and

VPS35 were blotted. Only VPS35 showed a specific enrichment with GST:RAB21. Gels were stained with coomasie blue to compare purified protein levels in all conditions, n=3 independent experiments.

Figure S2. TMED10 localization at trans-Golgi is modulated by RAB21.

(A) TMED10:GFP colocalization with endogenous TGN46 in parental and RAB21 knockoutHeLa cells. TMED10:GFP is stained in green, TGN46 in red and nucleus in blue.

Scale bar, 10μm. Dotted squares are magnified, scale bar 5 μm, n=3 independent experiments. (B) Same as in(A) immunofluorescence showing colocalization between endogenous ci-MPR andTMED10:GFP, n=3 independent experiments. (C) Quantification of

TMED10 and TGN46colocalization showed in (A). Histograms represent average Pearson correlation per cell, error barsare SEM. (D) Same as in (C) quantification of TMED10 and ci-MPR colocalization showed in(B), error bars are SEM. Mann-Whitney tests were used for statistical analysis in C and D.

161 162

Figure S3. TMED10 stability is not affected by blocking proteasomal or lysosomal functions.

Proteasome inhibition does not stabilize TMED10 over a 16h treatment. (A) A MG132 chase wasperformed to monitor TMED10 stability in parental and RAB21 gRNA-2 and -3 KO

HeLa celllines, n=3 independent experiments. Endogenous TMED10, RAB21, GAPDH and

Ubiquitin were assessed through western blotting. Note that the ubiquitin WB was performed on a different gelloaded with the same amount of proteins as for the GAPDH, RAB21 and

TMED10 immunoblots,and was used to confirm the efficiency of the MG132 treatment. (B)

A Bafilomycin A1 chase wasperformed to monitor TMED10 stability in parental and RAB21 gRNA-2 and -3 KO HeLa celllines, n=3 independent experiments. Endogenous TMED10,

RAB21, LC3-II and GAPDH wereassessed through western blotting. The accumulation of

LC3-II at 4h and 16h following BafA1treatment confirms that lysosomal functions were affected.

162 163

Article 2 Figure 1 : RAB21 interacts with TMED10

163 164

Article 2 Figure 2 : RAB21 interacts with TMED10

164

165

Article 2 Figure 3 : TMED10 Golgi localization is altered in RAB21 knockout cells

165

166

Article 2 Figure 4 : RAB21 knockout reduces TMED10 and TMED2 protein levels

166

167

Article 2 Figure S 1 Indirect interaction between RAB21 and TMED10

167 168

Article 2 Figure S 2 TMED10 localization at trans-Golgi is modulated by RAB21

168 169

Article 2 Figure S 3 TMED10 stability is not affected by blocking proteasomal or lysosomal functions

169 170

4DISCUSSION

Le trafic membranaire consiste en l’échange de lipides et de cargos protéiques entre les différents compartiments cellulaires via la formation et le transport de vésicules (Chavrier et Goud 1999). Ce processus est essentiel à l’homéostasie de toutes les cellules (Vicinanza et al. 2008). En effet, des défauts du trafic sont impliqués dans de nombreuses pathologies telles que les maladies neurodégénératives et le cancer. Ce trafic est hautement régulé par une large gamme de protéines qui contrôlent chacune des étapes de la sélection des cargos, la formation des vésicules, leur transport et la fusion de ces vésicules avec une membrane cible (Bhuin et Roy 2014). La coordination des protéines impliquées dans le trafic membranaire est donc un enjeu pour l’homéostasie cellulaire.

Les petites GTPases RAB sont considérées comme les chefs d’orchestre du trafic membranaire. En effet, les RAB recrutent des effecteurs (SNARE, SNX, protéines motrices, protéines d’ancrage, complexe de sélection des cargos) et régulent leur activité (Bhuin et Roy

2014). De la même manière que pour des défauts du trafic, une dérégulation de l’activité des

RAB peut aussi être directement impliquée dans le développement de pathologies

(Hutagalung et Novick 2011). En effet, des mutations activatrices dans le gène de RAB35 ont été identifiées comme oncogéniques et impliquées dans des cancers du col de l’utérus, des poumons et dans des lymphomes (Shaughnessy et Echard 2018). De même, l’amplification de l’expression de RAB1A a été caractérisée comme oncogénique et impliquée dans le développement du cancer colorectal (Thomas et al. 2014). Aussi, certaines

RAB sont surexprimées dans la maladie d’Alzheimer (RAB5, RAB7A, RAB4 et RAB27) ou

170

171 inactives dans la maladie de parkinson (RAB39B, RAB7A, RAB39B et RAB29) (Kiral et al.

2018).

La caractérisation des mécanismes d’action des RAB est donc un enjeu majeur à la fois pour la compréhension du trafic membranaire et pour l’identification de nouvelles cibles thérapeutiques. Prenons l’exemple de RAB21 qui est impliqué dans l’internalisation des intégrines et l’inhibition de l’anoïkose, parmi d’autres fonctions (Alanko et al. 2015).

L’identification des régulateurs et effecteurs de RAB21 impliqués dans cette voie permettrait de les cibler et d’inhiber leur action soit par des inhibiteurs pharmacologiques soit de peptides cycliques inhibant leur interaction. Plusieurs études d’identification des partenaires protéiques des RAB ont été réalisées mais se sont révélées inefficaces pour la caractérisation de certaines RAB dont RAB21 (Gillingham et al. 2014). Au vu de ses fonctions associées à la régulation de l’autophagie et des voies de signalisation EGF et FAK, et en considérant le manque concernant l’identification de ses effecteurs, RAB21 représentait un candidat très intéressant pour développer de nouvelles approches d’étude de protéomique (Alanko et al.

2015; Jean et al. 2015; Yang et al. 2012).

4.1 Comparaison des approches d’identification des partenaires de RAB21

L’interaction des effecteurs avec les RAB est régulée par les changements de conformation suite à l’échange du GTP en GDP ainsi que par leur localisation (Pylypenko et al. 2018). Les approches développées, à ce jour, consistent à identifier les effecteurs et les

GAP des RAB en caractérisant les partenaires protéiques des formes constitutivement actives des RAB. De même, les GEF ont été étudiées par l’étude des partenaires de la forme inactive des RAB. Ainsi, les études des RAB nécessitent l’expression des mutants Q-L, constitutivement actifs, et des mutants S-N, dominants négatifs (Gillingham et al. 2014).

171 172

Une première approche consiste à identifier par spectrométrie de masse l’ensemble des effecteurs des RAB et ce, suite à des purifications par affinité via l’utilisation de colonnes

GST (Gillingham et al. 2014). Cependant, les conditions de purifications, la difficulté à charger le GTP sur les RAB in vitro, ainsi que l’absence de contexte cellulaire ne permettent pas d’obtenir une vue d’ensemble des partenaires des RAB. Dernièrement, le développement de nouvelles approches de protéomique de marquage par proximité a offert de nouvelles possibilités. En effet, les approches BioID et APEX2 permettent de biotinyler in cellulo l’ensemble de l’environnement protéique d’une protéine d’intérêt, de purifier et de caractériser ces protéines par spectrométrie de masse (Roux et al. 2012; Lam et al. 2014).

Aussi, l’approche BioID a été récemment utilisée pour l’étude des RAB (Gillingham et al.

2019). Dans cette étude, les mutants Q-L et S-N ont été exprimées et relocalisées à la mitochondrie. Cette approche a permis d’identifier de nombreux effecteurs dans un contexte cellulaire. En revanche, l’utilisation de mutants et la relocalisation à la mitochondrie ne permet pas de rendre compte de la localisation endogène ni des variations attribuables aux changements de conformation des RAB. Enfin, ces approches ne permettent pas d’identifier des partenaires qui nécessitent la présence de plusieurs déterminants tels que des phospholipides (Jean et Kiger 2012).

Pour pallier ces limitations, nous avons choisi d’utiliser l’approche APEX2 avec la forme sauvage de RAB21. Afin de valider l’approche, nous avons comparé les résultats obtenus avec deux autres RAB localisées aux endosomes précoces, RAB4 et RAB5 et une

RAB associée aux endosomes tardifs, RAB7 (Hutagalung et Novick 2011). Tout d’abord, nous avons validé que l’induction de la biotinylation de chacune des RAB fusionnées à l’enzyme APEX2 soit bien sous le contrôle de l’ajout de H2O2 (Article 1, Figure 1C).

L’induction de la biotinylation semble enrichie aux endosomes précoces pour

172 173

APEX2:RAB4, 5 et 21 (Article 1, Figure 1D) et aux lysosomes pour APEX2:RAB7 (Article

1, Figure EV3E). Ces résultats suggèrent que la fusion des RAB avec APEX2 ne semble ni affecter l’activité d’APEX2, ni affecter la localisation des RAB. De plus, si l’on s’attarde sur la biotinylation induite par la protéine de fusion APEX2:RAB21, on constate un marquage présent sur des structures semblables au RE et, plus faiblement, diffus dans le cytosol (Article

1, Figure 1D). Cette biotinylation peut s’expliquer par le fait que RAB21 n’est pas seulement localisée aux endosomes précoces mais a aussi été identifiée au RE et au Golgi, dans certains contextes. En revanche, le marquage cytosolique observé peut être associé à une sous- population de RAB21 qui est sous sa forme inactive et donc solubilisée dans le cytosol via son interaction avec une GDI. Étant donné le contexte où APEX2:RAB21 est surexprimée, nous ne pouvons pas exclure qu’une partie du marquage observé soit la résultante d’une expression ectopique. Cependant, l’induction de biotinylation cytosolique ou réticulaire n’est pas observée avec la protéine de fusion APEX2:RAB7 (Article 1, Figure EV3E). Cela suggère que le marquage cytosolique est dépendant de la RAB et non de la fusion avec

APEX2.

Par ailleurs, les derniers résultats non publiés du laboratoire semblent démontrer que l’approche APEX2 est applicable à la plupart des RAB. En effet, cette approche s’est montrée efficace avec les RAB 1A, 2A, 3A, 4B, 5B, 6, 9, 11A, 18, 22A, 23, 25, 27A, 30, 32, 33B, et

35. La comparaison des protéines enrichies avec chacune des RAB montre que les RAB21,

5A et 4 forment une sous-classe. RAB7 semble être isolée. De même, RAB33 et RAB2 forme une autre sous-classe au sein des RAB. L’ensemble de ces résultats est cohérent avec les données de la littérature, et valide l’approche utilisée. De plus, les voies de signalisation associées au protéines enrichies avec chacune des RAB4, 5, 7 et 21 semblent cohérentes avec les fonctions attendues (Article 1, Figure 2B).

173 174

Lorsque l’on compare les enrichissements relatifs des protéines avec chacune des

RAB (Figure 2A et EV3D), on observe que RAB21 est très proche de RAB5, plus éloignée de RAB4 et très différente de RAB7. Ces résultats suggèrent que RAB5 et RAB21 possèdent des fonctions similaires et potentiellement redondantes. Cela confirme certaines données de la littérature. En effet, RAB5 et 21 sont toutes les deux associées aux endosomes précoces et régulent l’internalisation des intégrines (G. Stupack and A. Torres 2012; Pellinen et al. 2006).

La forte homologie entre RAB5 et RAB21 peut expliquer ces ressemblances. Malgré des similarités, il est possible de noter des spécificités avec chacune des RAB. En effet, le complexe EARP, effecteur nouvellement identifié pour RAB4, est identifié au complet et spécifique à RAB4 (Article 1, Figure 2C). Par ailleurs, le Rétromère est fortement enrichi avec RAB7 (Article 1, Figure2C). Enfin, SLC7A11, PDS5A, PIKFYVE, PIK3R2 et d’autres protéines sont enrichies uniquement avec RAB21 (Article 1, Figure EV4A et B). Ces différences observées confirment la notion de microdomaines générés par les RAB, et valident des spécificités fonctionnelles entre RAB5 et RAB21.

L’enzyme APEX2 permet de biotinyler l’environnement protéique mais ne permet pas de dissocier les interacteurs directs des protéines présentes dans l’environnement proche des RAB. Cela mène donc à l’identification d’un grand nombre de protéines. En effet, plus de 1300 protéines sont enrichies en combinant les quatre RAB. Il apparait évident que ces

1300 protéines ne sont pas des partenaires directs. Plus de mille protéines ont été identifiées avec RAB21 et 62 avec RAB7 (Article 1, Figure 2A). Cette différence peut s’expliquer par une densité accrue de cargos et d’effecteurs à la surface des endosomes précoces vis-à-vis de celles des lysosomes. Aussi, la localisation de RAB21 semble associée à divers compartiment, contrairement à RAB7 qui apparait restreint aux vésicules lysosomales. Il est tout de même nécessaire de considérer que les niveaux d’expression de RAB7 et RAB21

174 175 ainsi que des différences de disponibilité en biotin-phénol entre les lysosomes et les endosomes précoces peuvent expliquer les différences de niveaux de biotinylation.

L’identification de plus de 1000 partenaires potentiels ne permet pas, en soi, d’identifier les effecteurs et régulateurs spécifiques de chaque RAB. En revanche, lorsque l’on compare les enrichissements relatifs à chacune des RAB, il apparait de fortes spécificités.

Une autre méthode pour pallier ce problème consiste à combiner les résultats de l’approche APEX2 avec ceux de l’approche de protéomique quantitative SILAC, combinée

à la purification par affinité, appliquée aux protéines de fusion GFP:RAB21-WT, au dominant négatif GFP:RAB21-T33N et au constitutivement actif GFP:RAB21-Q78L.

Comme attendu, les protéines de fusion GFP:RAB21-WT et GFP:RAB21-Q78L sont essentiellement localisées aux endosomes précoces et GFP:RAB21-T33N au Golgi (Article

1, Figure EV1 A,B,C et D). Lorsque l’on compare les 272 protéines enrichies dans la lignée

HeLa et les 175 dans la lignée HCT116 (Article 1, Figure EV2A et B), on note des spécificités liées au type cellulaire. De plus, l’analyse des voies de signalisations enrichies montre qu’elles sont majoritairement liées à la traduction dans la lignée HeLa et aux mitochondries dans la lignée HCT116 (Article 1, Figure EV2D et E). Ces résultats sont éloignés des données de la littérature et des résultats obtenus avec l’approche APEX2. Ce n’est que lorsque l’on considère uniquement les protéines enrichies dans les deux lignées que l’on obtient un réseau de 29 protéines associées à des fonctions liées au trafic membranaire (Article 1, Figure EV2E,

F, et G). Les protéines identifiées semblent préférentiellement interagir avec la forme active de RAB21 (Article 1, Figure EV2C). Comparativement à l’approche APEX2, l’utilisation des protéines de fusion GFP:RAB21 nécessite la combinaison de multiples conditions afin d’obtenir des résultats cohérents. De plus, il semble qu’un grand nombre de partenaires soient perdus lors de la purification. Cependant, l’utilisation de l’approche APEX2 est limitée par

175 176 la présence de tyrosines accessibles afin d’être biotinylées. En effet, l’ensemble des cargos et effecteurs transmembranaires avec un domaine cytosolique ne présentant pas de tyrosine ne seront pas identifiés. Ici, l’approche de purification par affinité en contexte SILAC présente un avantage et permet l’identification de partenaires potentiels tels que TMED10,

SLC3A2 et Basiginin (Article 1, Figure EV1F). Ainsi, la combinaison des deux approches semble complémentaire et permit l’identification de nombreux cargos, effecteurs et régulateurs potentiels de RAB21 et ce, contrairement aux études réalisées précédemment.

4.2 RAB21 régule l’activité de WASH et du Rétromère

De manière intéressante, l’approche APEX2 a permis de mettre en évidence un lien entre RAB21 et les complexes WASH et Rétromère (Article 1, Figure 2C). En effet, le complexe WASH au complet et la protéine CAPZ, impliqués dans la polymérisation d’actine endosomale, sont enrichis seulement avec RAB21. Dans une moindre mesure, le complexe

WASH est enrichi avec RAB7 et plus faiblement avec RAB5. Par ailleurs, le CSC (VPS26-

29-35) du Rétromère, SNX27 et SNX3 sont enrichis majoritairement avec RAB7 mais aussi avec RAB5 et RAB21. Enfin, SNX1, 2,5 et 6 sont enrichis avec RAB21 et RAB5 et non avec

RAB7. RAB5 et RAB7 ont tous deux étés impliqués dans la régulation et le recrutement du

Rétromère aux endosomes précoces (Rojas et al. 2008; Priya et al. 2015). Ces données suggèrent que RAB7 serait impliquée dans les fonctions du Rétromère liées à SNX27 et

SNX3. Au contraire, RAB5 et RAB21 réguleraient potentiellement les fonctions du

Rétromère liées à SNX1/2 et 5/6. Ces données pourraient expliquer comment le Rétromère est impliqué dans des triages de cargos spécifiques depuis l’endosome précoce vers différentes destinations. De la même manière, le complexe WASH interagit majoritairement avec RAB21 vis-à-vis de RAB5 et RAB7. Or, les RAB peuvent agir en cascade dans la

176 177 régulation d’un même mécanisme (Jean and Kiger 2012). Il est donc possible que RAB21 recrute le complexe WASH au Rétromère suite au recrutement de ce dernier par RAB5 et/ou

RAB7. De plus, l’analyse de données bio-informatiques montrent que RAB7 et RAB21 coopèrent dans la régulation du complexe WASH et du complexe CCC qui possède des fonctions équivalentes au Rétromère (Clague and Urbé 2020). L’immunoprécipitation de protéines biotinylées dans les lignées APEX2:RAB21, APEX2RAB5 et APEX2:RAB4 confirme que le complexe WASH (Strumpelline) est majoritairement enrichi avec RAB21 et que le Rétromère (VPS35) l’est avec RAB5 (Article 1, Figure 3A). Les interactions entre

RAB21 et les complexes WASH et Rétromère ont été confirmées par des approches de co- immunuoprécipitation, de purification GST et de colocalisation par immunofluorescence

(Article 1, Figure 3A, B, C, D et E). Ces expériences confirment l’interaction entre RAB21 et le Rétromère ainsi qu’entre RAB21 et le complexe WASH. Cependant, elles ne permettent pas de définir la nature directe de ces interactions. FAM21, Strumpelline, et VPS35 interagissent majoritairement avec les formes sauvage et constitutivement active de RAB21 et aussi, mais en moindre mesure, avec le dominant négatif de RAB21. Le complexe WASH et le Rétromère ne peuvent donc pas être considérés comme des effecteurs de RAB21 à part entière, selon les définitions actuelles. Cependant, il a été montré que RAB11 interagit avec des effecteurs sous sa forme liée au GDP (Pylypenko et al. 2018). De plus, les dernières

études de BioID montrent que les dominants négatifs de RAB11 et de RAB5 interagissent avec des effecteurs potentiels et pas uniquement avec des GEF et des GDI (Gillingham et al.

2019). Ces résultats amènent une nouvelle question, à savoir comment définir les effecteurs et l’activité potentielle des RAB sous leurs formes liées au GDP ou au GTP. En effet, le changement de conformation induit par l’échange du GTP en GDP peut déstabiliser et non inhiber complètement l’interaction entre une RAB et ses effecteurs (Pylypenko et al. 2018).

177 178

De la même manière, certaines RAB subissent des changements de conformation plus ou moins drastiques pouvant expliquer que certaines interactions perdurent malgré l’échange de

GTP par le GDP. De plus, il est possible que ce changement de conformation induise l’interaction avec de nouveaux effecteurs et pas seulement avec des GAPs et des GDI.

Par ailleurs, RAB21 colocalise fortement avec FAM21 et VPS26 et en moindre mesure avec WASH (Article 1, Figure 3E). Il est donc possible que RAB21 recrute le complexe WASH au Rétromère via l’interaction avec FAM21 qui fait partie du complexe

WASH et interagit avec VPS35. De la même manière, FAM21 permet le recrutement de

CAPZ (Park et al. 2013). Aussi, il est possible que RAB21 soit impliquée dans l’assemblage du complexe WASH à l’endosome. La déplétion de RAB21 par l’approche CRISPR/Cas9 affecte les niveaux d’expression des membres du CSC mais pas ceux des membres du complexe WASH (Article 1, Figures 5A, B, et C). Or, les niveaux d’expression des membres du complexe WASH sont interdépendants. Il est donc peu probable que RAB21 soit impliquée dans l’assemblage du complexe WASH. De plus, la déplétion de RAB21 n’affecte que partiellement le recrutement du complexe WASH aux endosomes précoces (Article 1,

Figures 5F, G et H). Il est possible qu’en l’absence de RAB21, le complexe WASH soit recruté par la sous-population du complexe du Rétromère toujours présente aux endosomes précoces ainsi que la protéine HRS. Par ailleurs, le rôle de RAB21 dans le recrutement du complexe WASH pourrait être compenser par d’autres RAB endosomales. En revanche, la déplétion de RAB21 inhibe fortement les niveaux de polymérisation de F-actine aux endosomes précoces (Article 1, Figures 5I et J). RAB21 est donc essentielle à l’activité endosomale de WASH plus qu’à son recrutement.

178 179

La délétion de RAB21 semble affecter le recrutement du CSC aux endosomes précoces (Article 1, Figure 5C, D, et E). Cependant, il est difficile de discriminer les défauts de localisation aux endosomes précoces des défauts d’expression du Rétromère. Au contraire, la déplétion de RAB21 n’affecte pas les niveaux d’expression de SNX1 mais inhibe partiellement son recrutement aux endosomes précoces (Article 1, Figure EV5 A, B et C).

Le maintien du recrutement du Rétromère et du complexe WASH aux endosomes précoces, et ce en absence de RAB21, peut être expliqué par l’activité de RAB7 et de SNX3. De même, la protéine HRS associée à la dégradation lysosomale médie le recrutement de WASH aux endosomes précoces (MacDonald et al. 2018). Il est donc possible que RAB21 ne soit pas impliquée dans le recrutement du complexe WASH aux endosomes précoces mais plus spécifiquement dans la coordination de l’activité de WASH et des complexes de sélection des cargos. En effet, il a déjà été montré que RAB5 et RAB7 agissent successivement pour coordonner l’activité des SNX1/2-5/6 et du CSC du Rétromère (Rojas et al. 2008). RAB21 pourrait agir dans la continuité de cette cascade afin de médier l’activité de WASH. Aussi, il est possible que RAB21 soit impliquée dans le recrutement d’une sous-population de WASH et du Rétromère. En effet, la déplétion de RAB21 semble impliquée dans le recyclage à la membrane plasmique d’une sous-catégorie de cargos, dits clathrine-indépendants.

La délétion de RAB21 affecte le recyclage à la membrane plasmique via la formation de tubules de recyclage des cargos SLC3A2, CD44, et de la Basigin (Article 1, Figures 6G et H). Il est intéressant de noter que SLC3A2 et Basigin ont été identifiés comme des cargos potentiels de RAB21 via l’approche AP-MS (Article 1, Figure EV1F). Au contraire, la déplétion de RAB21 n’affecte pas le transport rétrograde au réseau transgolgien des cargos

CI-MPR et GLUT1 (Article 1, Figure 6A et B). De plus, l’absence de RAB21 induit la dégradation lysosomale de SLC3A2 (Article 1, Figures EV5F, G, H et I). Il est à noter que

179 180 des défauts de recyclage de cargos sont régulièrement associés à leur dégradation. Ici, la déplétion de RAB21 semble donc orienter le trafic des cargos vers la dégradation. Cela peut

être un effet direct de l’inhibition de l’activité de WASH. En effet, la polymérisation d’actine aux endosomes précoces est essentielle à la formation de tubules mais permet aussi d’inhiber la voie de dégradation en générant des frontières d’actine qui délimitent physiquement des régions de recyclage. De plus, le Rétromère inhibe la voie de dégradation par le recrutement de TBC1D5 et RME-8 qui inhibent respectivement RAB7 et ESCRT (Seaman et al. 2009;

Norris et al. 2017).

Une autre possibilité expliquant les défauts de polymérisation d’actine serait que

RAB21 régule plus directement l’activité de WASH. En effet, le complexe WASH est activé par l’ubiquitination de WASH1 par l’ubiquitine ligase TRIM27, ou l’un des partenaires potentiels spécifiquement enrichi avec RAB21 par l’approche APEX2 est USP7 qui déubiquitine WASH1 (Article 1, FigureEV4B). Il a été montré que le cycle d’ubiquitination et de déubiquitination de WASH était essentiel dans l’induction de la polymérisation d’actine. De plus, TRIM33, une ubiquitine ligase, est spécifiquement enrichie avec RAB21.

Il n’est donc pas impossible que RAB21 régule l’activité du complexe WASH en induisant son ubiquitinylation via le recrutement d’ubiquitines ligases autres que TRIM27.

Enfin, les données des analyses d’APEX2:RAB21 montrent des enrichissements spécifiques de C16orf62 membre du complexe Retriever, et de certains membre du complexe

CCC. Des études d’analyses bioninformatiques confirment le rôle de RAB21 dans la régulation de l’activité du complexe CCC (Clague et Urbé 2020). En effet, dans cette étude, il a été montré par l’analyse de cribles CRISPR/Cas9 que RAB21 et les membres COMMD2,

6, et 8 du complexe CCC sont co-dépendants. Par ailleurs, le complexe CCC qui est lié au

Retriever, inhibe l’activité endosomale de WASH, et réduit l’accumulation de PI(3)P par le

180 181 recrutement de la phosphatase MTMR2 (Singla et al. 2019). En revanche, la déplétion de

RAB21 semble favoriser le recrutement de C16orf62 aux endosomes précoces. De plus,

MTMR13 qui interagit directement avec MTMR2 est une GEF de RAB21. Enfin, les derniers résultats du laboratoire semblent montrer que la déplétion de RAB21 induit une baisse de l’accumulation du PI(3)P aux endosomes précoces. RAB21 et le complexe CCC semblent avoir des fonctions antinomiques. Ici, RAB21 semble donc moduler la balance entre le

Rétromère et le Retriever afin d’influencer le type de cargos sélectionnés.

La GEF VARP et la GAP TBC1D5 lient toutes les deux le même domaine de VPS29 et il a été montré que la déplétion de VARP inhibe le recyclage de MCT-1. Confirmant les données de la littérature, VARP est exclusivement enrichie avec RAB21 (Article 1, Figure

2C). De la même manière, la déplétion de RAB21 inhibe fortement la présence de MCT-1 à la membrane plasmique (Article 1, Figures 6D et E). Il est donc fort probable que RAB21 soit associée au Rétromère en présence de VARP régulant ainsi le recyclage à la membrane d’une sous-catégorie de cargos clathrine-indépendants. L’ensemble des résultats décrits dans cette études sont des observations et il sera nécessaire de réaliser d’autres études afin de décrire les mécanismes moléculaires de régulation de l’activité des complexes WASH et

Rétromère par RAB21. Aussi, de plus amples études permettront de définir comment RAB5,

7 et 21 sont coordonnées afin de réguler ces complexes. Cependant, il est très intéressant de noter que la simple comparaison des différences d’enrichissements protéiques avec chacune des RAB permet de suggérer l’implication et la coordination des RAB5, 7 et 21 dans la régulation des processus de triage dépendants du Rétromère.

181 182

4.3 Rôle de RAB21 dans la régulation de TMED10

Les résultats de spectrométrie de masse quantitative via l’approche AP-MS en condition SILAC montrent une interaction entre RAB21 et TMED10/TMED9. L’interaction entre RAB21 et TMED10 a été confirmée par des co-immunoprécipitations dans les lignées

HeLa et HCT116 suggérant que cette interaction soit conservée à travers les types cellulaires

(Article 2, Figure 1A et B). Par ailleurs, les données de spectrométrie montrent que TMED10 et TMED9 interagissent préférentiellement avec les formes actives et sauvages de RAB21.

Ce résultat a été confirmé par des expériences de PLA (Proximity Ligation Assay) (Article

2, Figures 2A, B, et C). Les données de spectrométrie semblent aussi montrer une interaction spécifique de TMED10 avec les formes sauvage et constitutivement actives de RAB21

(Article 2, Figure S1A et 2C). Cela pourrait suggérer que TMED10 est un effecteur de

RAB21. Cependant les expériences de purification GST ne montrent pas d’interaction de

RAB21 avec TMED10. N’interagissant pas directement ensemble, il est peu probable que

TMED10 soit un effecteur de RAB21.

En revanche, il est possible que TMED10 soit un cargo dont le triage est régulé par

RAB21. TMED10 est une protéine de la famille p24 essentiellement impliquée dans la formation des vésicules COPI et COPII. TMED10 cycle donc entre le Golgi et le RE. De manière intéressante, nos données montrent que la déplétion de RAB21 affecte la localisation au cis-Golgi et au réseau transgolgien de TMED10 ( Article 2, Figures 1A, 1B, S2A et SC).

RAB21 est majoritairement associée aux endosomes précoces mais des données de littérature montrent que la GEF VARP active RAB21 au Golgi et régule le triage de VAMP7.

N’interagissant pas directement avec TMED10, il est donc possible que RAB21 soit impliquée dans la rétention de TMED10 entre le Golgi et le RE via la régulation d’un

182 183 effecteur. Aussi, des données montrent que l’interaction de TMED10 avec certains cargos permet la séquestration de ces derniers au Golgi. Dans ce cas, RAB21 pourrait favoriser l’interaction de TMED10 avec un cargo ou induire le trafic d’un cargo vers le Golgi. Par ailleurs, il a été montré qu’une sous-population de TMED10 est présente aux vésicules de sécrétion et à la membrane plasmique. En effet, TMED10 permet la sécrétion de protéines associées à une ancre GPI. De plus, les derniers résultats de la littérature montrent que

TMED10 permet le recyclage au Golgi depuis la membrane plasmique de cargos possédant une ancre GPI avec des défauts de conformations (Pastor-Cantizano et al. 2016). Nous avons vu précédemment que la déplétion de RAB21 n’affecte pas le transport rétrograde de CI-

MPR. Cependant, il a été montré que le transport rétrograde de cargos régulé par le CSC de l’endosome précoce vers le Golgi dépendait des golgines (Cui et al. 2019). De plus, il a été montré que TBC1D23 qui est spécifiquement enrichi avec RAB21, interagit avec WASH, et les golgin-97 et golgine-245. Il est donc possible qu’un effecteur de RAB21 tel qu’une golgine soit impliqué dans le transport rétrograde d’un certain type de cargos dont TMED10.

Par ailleurs, RAB21 régule l’endocytose de nombreux cargos tels que les intégrines et le récepteur à l’EGF. RAB21 pourrait donc être impliquée dans l’endocytose de cargos associés

à une ancre GPI.

La délétion de RAB21 induit une baisse des niveaux d’expression de TMED10 et de

TMED2 (Article 2, Figures 4A, B et C). Cela confirme les données de la littérature qui montrent que l’expression de membres de la famille p24 est interdépendante. En revanche, l’analyse des cinétiques de dégradation protéique ne semble pas montrer d’effet sur la stabilité de TMED10 lors de la déplétion de RAB21 (Article 2, Figures 4F et G). Les données de la littérature montrent que TMED10 a une demi-vie de trois heures dans des cellules neuronales et est dégradée par le protéasome. En revanche, TMED2 a une demi-vie beaucoup

183 184 plus longue dans les cellules Vero (Pastor-Cantizano et al. 2016). Nos données suggèrent que la demie de TMED10 est beaucoup plus longue dans les cellules HeLa. En effet, on n’observe pas de diminution des niveaux d’expression suivant les sept premières heures de l’inhibition de la synthèse protéique. Les mécanismes de dégradation de TMED10 ne semblent donc pas conservés d’une lignée cellulaire à l’autre. La faible demi-vie de TMED10 dans les neurones peut s’expliquer par la forte activité sécrétrice de ce type cellulaire. De plus, l’inhibition du protéasome et des lysosomes pendant 16 heures ne semble pas affecter, non plus, les niveaux d’expression de TMED10. Enfin, la déplétion de RAB21 n’affecte pas la transcription des gènes de TMED10 et TMED2. RAB21 pourrait affecter l’expression d’une autre protéine de la famille p24 ce qui aurait pour effet de réduire les niveaux d’expression de TMED2 et

TMED10 sur le long terme.

Finalement, TMED10 interagit directement avec la y-sécrétase et inhibe son activité catalytique. De la même manière, RAB21 interagit avec la préséniline 1, induit son triage vers la dégradation lysosomale inhibant le clivage d’APP par la -sécrétase (Sun et al. 2017).

Ces données suggèrent une implication directe et indirecte de RAB21 dans le développement de la maladie dAalzheimer. De plus, TMED10 est impliquée dans la régulation des étapes précoces de l’autophagie via l’inhibition d’ATG4 par interaction directe (Shin et al. 2019) puiqu’ATG4 permet la synthèse de LC3-II. RAB21 est impliquée dans la fusion des autophagosomes avec les lysosomes et donc dans la régulation des étapes tardives de l’autophagie. Le rôle potentiel de RAB21 dans la régulation de l’autophagie via TMED10 suggère un impact de RAB21 dans la modulation de plusieurs étapes du flux autophagique.

184 185

4.4 Nouvelles fonctions potentielles de RAB21

La combinaison des approches de spectrométrie de masse a permis d’identifier deux nouvelles fonctions pour RAB21 dans le recyclage de cargos clathrine-indépendants et dans la régulation de la localisation de TMED10. En revanche, de nombreux autres partenaires potentiels suggèrent des fonctions qui n’ont pas été étudiées ici. De plus, certains cargos et régulateurs de RAB5 et RAB21, à savoir les intégrines 1 et 3, et RABGEF1 ont été identifiées par l’approche APEX2 comme étant majoritairement enrichie avec RAB4

(Article 1, Figure 2C). Cela ne remet pas en question la validité des résultats car il a été montré qu’un même cargo peut être trié par différentes RAB. De même, une GEF peut aussi activer différentes RAB. Au contraire, cela suggère un nouveau rôle potentiel pour RAB4 dans le trafic des intégrines 1.

De nombreuses protéines sont enrichies spécifiquement avec RAB21, dont certaines sont impliquées dans le métabolisme des phopholipides, ce qui est essentiel au maintien de l’identité membranaire et à la régulation du trafic (Article 1, Figure EV4A). En effet,

MTMR3, 6, 9, et 12 sont des phosphatases impliquées dans la synthèse de PI(3)P et sont spécifiquement enrichies avec RAB21 (Robinson et Dixon 2006). Or, le PI(3)P est particulièrement enrichie aux endosomes précoces. Enfin, les données de la littérature montrent que la GEF MTMR13 de RAB21 régule l’hydrolyse du PI(3)P en phosphatidylinositol via le recrutement de MTMR2. Aussi, MTMR13 régule la synthèse de

PI(3)P par l’interaction avec PI3KC2 (Jean et al. 2012) Ainsi, RAB21 semble être activement impliquée dans la régulation du métabolisme du PI(3)P endosomal.

De plus, RAB21 est spécifiquement associée aux protéines PLEKHM2 (Pleckstrin

Homology And RUN Domain Containing M2), et SARA (Smad Anchor for Receptor

185 186

Activation) (Article 1, Figure EV4 B). L’interaction entre RAB21 et ces deux protéines a été confirmée par des expériences de co-immunoprécipitation (Article 1, Figures EV4C et D).

La protéine SARA régule l’activation de la voie TGF- aux endosomes précoces (Di

Guglielmo et al. 2003). Or, il a été montré que l’activité de voie de signalisation aux endosomes précoces est dépendante de la polymérisation d’actine qui permet de générer des plateformes de signalisation qui stabilisent la présence de complexes de signalisation. De même, RAB21 est impliquée dans la signalisation de FAK aux endosomes (Alanko et al.

2015). Cette régulation a été associée au rôle de RAB21 dans l’endocytose des intégrines 1.

Il est donc possible que le recrutement du complexe WASH par RAB21 soit directement impliqué dans le maintien de l’activité de la voie FAK aux endosomes précoces et joue un rôle dans la régulation de la voie de signalisation TGF-. PLEKHM2 est une protéine motrice permettant le transport des lysosomes le long des microtubules (Rosa-Ferreira and Munro

2011). Par ailleurs, RAB21 régule le triage de VAMP8 vers les lysosomes. Dans ce contexte

PLEKHM2 pourrait agir comme un effecteur régulant le triage de VAMP8 vers les lysosomes.

Aussi, il a été montré que la position des lysosomes dans la cellule influence l’activité de mTOR (mammalian Target Of Rapamicin) (Korolchuk et al. 2011). En effet, les lysosomes situés dans des régions périnucléaires sont associés à l’activité de mTORC1

(Figure 11). Au contraire, les lysosomes proches de la membrane plasmique sont associés à la signalisation de mTORC2 (Jia and Bonifacino 2019). Par ailleurs, les études de spectrométrie de masse quantitatives montrent que RAB21 interagit avec mTOR et LST8.

Ces deux protéines sont des membres communs aux complexes mTORC1 et mTORC2 (Kim et al. 2017). De plus, il a été montré que la délétion du Rétromère favorise l’activation de

186 187

RAB7, ce qui inhibe l’activité de mTORC1 aux lysosomes. Cette délétion des membres du

Rétromère a aussi pour effet de favoriser l’activité de mTORC2 (Kvainickas and Heike

2019). Les derniers résultats du laboratoire, non présentés ici, confirment l’interaction entre

RAB21 et mTOR. De plus, cette interaction est dépendante de l’état de carence. Au contraire des effets de la délétion du Rétromère, la délétion de RAB21 inhibe l’activité de mTORC2 et favorise l’accumulation de la protéine mTOR au lysosome (Figure 11).

Figure 11 : Rôle potentiel de RAB21 dans la régulation mTOR La signalisation mTOR est dissociée en deux voies de signalisation associées à deux complexes, mTORC1, et mTORC2. Le complexe mTORC1 est majoritairement associé aux lysosomes et activé par la concentration en acides aminés intracellulaires. mTORC1 induit la phosphorylation de la S6K et la 4eBP1. Le complexe mTORC2 est localisé principalement à la membrane plasmique et est activé en réponse à des facteurs de croissance. L’activation de mTORC2 induit la phosphorylation de la sérine 473 d’AKT. La déplétion de RAB21 induit une accumulation de la kinase mTOR au lysosome ayant pour effet potentiel de dépléter le réservoir de mTOR à la membrane plasmique et donc d’inhiber la voie de signalisation mTORC2, ce qui se traduit par une baisse de la phosphorylation de AKTser473.

De même, la délétion de MTMR13 inhibe l’activité de mTORC2 et la délétion de

VARP favorise l’activité de mTORC2. MTMR13 régulerait ainsi le triage induit par RAB21 vers les lysosomes et les résultats présentés ici suggèrent que VARP régule le triage médié par RAB21 vers le recyclage à la membrane plasmique. Il est donc possible qu’en fonction

187 188 du stimulus RAB21 soit activée par différentes GEF qui favorisent le triage via le Rétromère soit vers la membrane plasmique soit vers les lysosomes. Aussi, le triage du Rétromère par

RAB21 vers la membrane plasmique dépléterait le réservoir de Rétromère présent aux lysosomes et favoriserait l’inhibition et le reflux de mTORC1. Ainsi, RAB21 serait impliquée dans la composition des différents réservoirs cellulaires de mTOR ayant un impact direct sur ses niveaux d’activité.

Si l’on analyse la localisation de chacune des 29 protéines composant l’interactome de RAB21 par l’approche de spectrométrie quantitative, on constate que RAB21 est potentiellement associée à différents compartiments cellulaires. Premièrement, plusieurs transporteurs présents à la membrane plasmique sont enrichis avec RAB21. Deux possibilités peuvent expliquer ces résultats. Soit RAB21 régule leur recyclage depuis les endosomes précoces vers la membrane plasmique, comme cela est le cas pour SLC3A2 et Basiginin. Une autre possibilité serait que RAB21 régule leur endocytose comme cela est le cas pour les intégrines et le récepteur à l’EGF.

Par ailleurs, RAB21 interagit potentiellement avec des protéines cytosoliques. La

GDI2 est une GDI régulant la solubilisation de RAB21. GCN1 est une protéine impliquée dans la traduction. Le complexe mTORC1 régule directement l’activité de la machinerie de traduction (Marton et al. 1997). Ainsi, GCN1 pourrait être enrichie via l’interaction de

RAB21 avec mTOR qui serait présent au niveau de la machinerie de traduction. Pour sa part,

PYGB est une enzyme cytosolique qui est aussi associée aux lysosomes (Mathieu et al.

2016). De la même manière que pour VAMP8, il est possible que RAB21 régule le triage de

PYGB vers les lysosomes. Enfin, la protéine CFL1 est impliquée dans la polymérisation de

F-actine et de G-actine (Kanellos et al. 2015). CFL1 pourrait être un nouvel effecteur de

RAB21 impliqué dans la polymérisation de l’actine endosomale.

188 189

Deux protéines associées au noyau, XPO1 et CSEL1, ont été identifiées. Il a été montré que le complexe WASH était présent dans le noyau et y régule l’organisation de la chromatine par l’interaction avec la lamine A (Verboon et al. 2015). Par ailleurs, les images de microscopie confocale suggèrent que la délétion de RAB21 semble sensiblement altérer la morphologie du noyau. En effet, il est possible d’observer des surfaces irrégulières dans les colorations au DAPI des cellules délétés pour RAB21. Il est possible que la présence du complexe WASH au noyau soit en partie dépendante de l’activité de RAB21.

Trois protéines associées à la mitochondrie sont aussi présentes dans ce réseau. Le

Rétromère est impliqué dans la régulation de la mitophagie (Jimenez‐Orgaz et al. 2018). Ce rôle du Rétromère dans la régulation de la mitophagie et la relation entre RAB21 et le

Rétromère pourrait expliquer pourquoi des protéines mitochondriales sont associées avec

RAB21.

Enfin, une dizaine de protéines associées au RE et au Golgi semblent associées à

RAB21. La relation entre RAB21 et les protéines TMED9 et 10 a été discutée précédemment.

SURF4 (Surfeit 4) interagit avec TMED10 et est impliquée dans la régulation de la synthèse des vésicules COPI (Mitrovic et al. 2008). SURF4 pourrait être un effecteur de RAB21 impliqué dans la régulation de la localisation de TMED10. Les protéines NAPA (NSF

Attachment Protein Alpha) et NSF (N-Ethylmaleimide Sensitive Factor) sont des protéines respectivement impliquées dans le transport et l’ancrage des vésicules entre le RE et le Golgi

(Beckers et al. 1989). Ces protéines sont toutes impliquées dans des fonctions similaires. Il est donc difficile de confirmer si l’effet observé de RAB21 sur la localisation de TMED10 est médié par ces autres partenaires ou inversement. Enfin, il a été montré que les points de contact entre les endosomes précoces et le réticulum au niveau des points de polymérisation

189 190 d’actine sont impliqués dans la fission des tubules de recyclage. Ces protéines résidentes du

RE pourraient être impliquées avec RAB21 dans la régulation de ces évènements de clivage.

Selon le contexte cellulaire et son état d’activation, RAB21 a été identifiée aux endosomes précoces, au RE au Golgi et à la membrane plasmique. L’ensemble des résultats de spectrométrie de masse confirme la présence de RAB21 à ces différents compartiments.

Les effets de RAB21 aux endosomes, au Golgi, et au RE peuvent en partie être expliqués par sa relation fonctionnelle avec les complexes WASH et Rétromère. En revanche, cela ne permet pas d’expliquer comment RAB21 régule l’endocytose de nombreux cargos depuis la membrane plasmique. Au vu des différentes fonctions associées à RAB21 et en considérant que les complexes WASH et Rétromère soient impliqués dans la régulation de chacune de ces fonctions, la composition de l’environnement protéique de RAB21 (effecteurs, GEF et

GAP) devrait varier selon le contexte. Ainsi, des protéines régulatrices spécifiques à certains stimuli réguleraient chacune des fonctions associées à RAB21.

Il a précédemment été discuté les rôles potentiels de VARP et MTMR13 dans la régulation de RAB21. Par ailleurs, la GEF GAPVD1, spécifique à RAB5, est aussi enrichie avec RAB21. Par ailleurs, aucune GAP n’a été identifiée pour inhiber RAB21. Les données d’APEX2 montrent que TBC1D4, TBC1D15, GAPCENA, TBC1D5 et TBC1D23 sont enrichies avec RAB21. TBC1D5 et TBC1D23 sont associées avec le Rétromère et le complexe WASH (Kvainickas and Heike 2019; J. J. H. Shin et al. 2017). Il serait intéressant d’étudier si ces GAP sont spécifiques à RAB21 ou si elles sont des effecteurs de RAB21 associées au Rétromère et à WASH. En effet, certaines GAP sont des effecteurs de RAB inhibant d’autres RAB permettant l’action en cascades de plusieurs RAB impliquées dans un même processus.

190 191

5CONCLUSION ET PERSPECTIVES

RAB21 est une petite GTPase impliquée dans des évènements de triage associés au trafic endosomal. Plusieurs études ont démontré que RAB21 est directement impliquée dans la régulation de l’autophagie, et de la signalisation EGF et FAK. En effet, des carences en nutriments induisent l’activation de RAB21 par la GEF MTMR13 qui induit le triage de

VAMP8 vers les lysosomes et médie la fusion entre l’autophagosome et le lysosome. Lors de traitement de cellules à l’EGF, RAB21 induit l’internalisation de l’EGFR et son triage vers les lysosomes, ce qui a pour effet d’inhiber la signalisation EGF. Enfin, RAB21 permet l’internalisation des intégrines à la suite de leur activation, et induit l’activation de la voie

FAK, liée aux intégrines, à la surface de endosomes précoces. De plus, la déplétion de

RAB21 inhibe le potentiel de cellules cancéreuses à former des métastases lorsqu’elles sont injectées dans des souris. L’ensemble de ces données suggèrent une régulation différentielle selon les stimuli de RAB21 et une implication dans des processus pro-oncogéniques.

Cependant, l’ensemble des études protéomiques réalisées à ce jour ont échoué à définir les effecteurs de RAB21 impliqués dans chacune de ces fonctions. Cette thèse visait à utiliser les dernières approches de spectrométrie de masse afin d’identifier les effecteurs et régulateurs de RAB21. La combinaison des approches de purification par affinité en SILAC et APEX2 s’est montrée très efficace pour identifier les partenaires protéiques de RAB21.

L’approche quantitative SILAC a permis d’identifier une trentaine de partenaires protéiques pour RAB21. Parmi ces 29 protéines, les expériences de coimmunoprécipitations réalisées dans les cellules HeLa ont confirmé l’interaction de RAB21 avec TMED10, mTOR et SLC3A2. Enfin, les études fonctionnelles réalisées ont démontré un rôle pour RAB21 dans

191

192 le recyclage à la membrane plasmique de SLC3A2 et de Basigin. Il a aussi été démontré que

RAB21 est impliquée dans la localisation de TMED10 au cis-Golgi. Enfin, RAB21 semble médier la relocalisation de mTOR à la membrane plasmique et réguler la balance entre l’activité mTORC1 et mTORC2. Ainsi, les expériences fonctionnelles et de coimmunoprécipitation ont montré un lien entre RAB21 et cinq des protéines parmi les 29 du réseau généré. L’approche de purification par affinité en conditions quantitatives s’est donc révélée une solution fiable pour cibler et identifier de nouvelles fonctions de RAB21.

Toutefois, l’obtention de ces résultats a nécessité l’utilisation des trois formes de RAB21 et ce, dans deux lignées cellulaires. Cela représente donc un processus relativement lourd.

Enfin, cette approche permis d’identifier essentiellement des cargos potentiels pour RAB21, peu d’effecteurs liés au trafic membranaire, et aucune GAP ou GEF potentielles. Il est donc

évident que les conditions de purification ne permettent pas de rendre compte de l’intégralité de partenaires de RAB21.

La nature même de l’approche APEX2 de marquage par proximité s’affranchie des difficultés de purification des partenaires protéiques de RAB21. En effet, l’application de l’approche APEX2 à RAB21 a permis d’identifier de nombreuses GEF, GAP et effecteurs potentiels. L’approche APEX2, en utilisant uniquement la forme sauvage de RAB21 et ce, dans une seule lignée, permit de générer un interactome présentant la plupart des partenaires connus de RAB21 à savoir les intégrines 1 et 3, VARP, RABEX, MTMR2 qui est constitutivement liée à MTMR13, l’EGFR, et VAMP7. Par ailleurs, cette approche s’est montrée hautement efficace pour de nombreuses RAB. Ici, une seule expérience permit d’identifier des complexes au complet étant connus pour être associés à certaines RAB. Ainsi, le complexe EARP a été identifié comme partenaire de RAB4 et le complexe du Rétromère

192 193 avec RAB7. L’inconvénient de cette approche est le nombre de protéines enrichies avec chaque RAB. Par exemple, plus de 1000 sont identifiées pour RAB21. C’est la comparaison des enrichissements relatifs à chaque RAB qui permet de mettre en évidence des effecteurs et régulateurs spécifiques. De cette manière, il a été identifié une relation entre RAB21 et les complexes WASH et le Rétromère. Par la suite, les expériences de coimmunoprécipitation et de déplétion de RAB21 ont démontré que ces deux complexes sont des effecteurs de RAB21 qui médient son implication dans le recyclage de cargos clathrine-indépendants vers la membrane plasmique.

L’ensemble de ces expériences ont prouvé l’efficacité de l’approche APEX2 dans l’étude des RAB. L’application de cette approche à l’étude de RAB21 permit de démontrer son rôle dans le recyclage de cargos clathrine-indépendants et dans la régulation d’évènements de triages liés au Golgi. On peut donc définir 3 champs d’action de RAB21, à savoir l’internalisation des cargos depuis la membrane plasmique, le triage de cargos à l’endosome vers la membrane plasmique et vers le lysosome, et enfin un rôle dans la régulation d’évènements au Golgi.

Il serait donc intéressant de définir les GAP, GEF et effecteurs impliqués dans chaque fonction de RAB21. Aussi, l’approche APEX2 présente l’avantage d’induire la biotinylation de l’environnement protéique de RAB21 sur des temps très courts, de l’ordre de la minute.

La combinaison de l’approche APEX2 avec l’approche SILAC permettrait de définir les variations de l’environnement protéique de RAB21 induites par des stimuli. La stimulation de cellules avec l’EGF ou le réensemencement sur collagène permettrait de définir les effecteurs de RAB21 impliqués dans l’internalisation de cargos depuis la membrane plasmique. Aussi, des carences en nutriment définiraient les effecteurs impliqués dans le triage médié par RAB21 vers le lysosome. Par ailleurs, la comparaison de ces résultats

193 194 validerait si le triage vers le lysosome de l’EGFR et de VAMP8 par RAB21 implique les mêmes effecteurs. Cela permettrait aussi de caractériser l’implication des complexes WASH et Rétromère dans les différentes fonctions de RAB21.

Un autre axe de recherche intéressant serait de déterminer comment RAB5, 7 et 21 sont coordonnées dans la régulation du triage endosomal via l’action du Rétromère et du complexe WASH. En effet, les données de la littérature, et les résultats de protéomique via l’approche APEX2 montrent que les fonctions de ces trois RAB sont liées à ces deux complexes. L’étude des RAB a montré qu’elles pouvaient agir en cascade dans une même voie du trafic membranaire. Il serait donc intéressant de réaliser des études d’épistasie. Cela permettrait de définir comment RAB5, 7 et 21 régulent WASH et le Rétromère, et de définir si elles agissent en cascade ou compétitionnent entre elles pour recruter ces deux complexes et ainsi médier différents évènements de triage.

Enfin, RAB21 a été montrée pour jouer un rôle dans la formation de métastases via l’activation de la voie FAK et l’inhibition de l’anoïkose lorsque les cellules sont en suspension. Or, RAB21 est impliquée dans la régulation de l’autophagie et semble jouer un rôle dans la régulation de la voie mTOR. Il serait donc intéressant de définir si l’effet de

RAB21 observé sur la survie des cellules cancéreuses est dépendant uniquement de son rôle dans la régulation de la voie FAK ou si l’autophagie et la voie mTOR sont aussi impliquées.

Afin de répondre à cette question il serait intéressant de dépléter par l’approche

CRISPR/Cas9 les effecteurs spécifiques à chacune des voies de RAB21 préalablement identifiées et de mesurer leur rôle dans la régulation de la survie des cellules en absence d’ancrage. Enfin, la polymérisation d’actine peut jouer un rôle dans l’activation de voies de signalisation à la membrane des endosomes. Il serait donc intéressant d’étudier l’effet de la déplétion des différents membres du complexe WASH sur les niveaux de phosphorylation

194 195 de FAK et ce, à la suite d’un réensemencement sur collagène. Pour finir, des mutations dans les gènes des membres des complexes WASH et Rétromère ainsi que l’inhibition de l’expression de TMED10, et des défauts d’autophagie sont tous associés à des pathologies neurodégénératives. Ces données suggèrent fortement un rôle potentiel pour RAB21 dans le développement de maladies neurodégénératives qu’il serait intéressant d’étudier. La poursuite de l’étude des partenaires spécifiques à chacune des voies associées à RAB21 pourrait permettre d’identifier des cibles thérapeutiques prometteuses dans la lutte contre les maladies neurodégénératives et le cancer.

195 196

6REMERCIEMENTS

Je remercie tout particulièrement le Pr. Steve Jean, mon directeur de recherche qui m’a offert l’opportunité d’entreprendre un doctorat à un moment où cela n’était plus une évidence. Je le remercie aussi pour son encadrement, sa patience et sa capacité à orienter les projets sans les cloisonner et laisser la possibilité à chacun d’être maître de ses travaux. Je le remercie également pour avoir su façonner une super équipe et un environnement de travail agréable.

Je remercie chacun des membres du laboratoire Steve Jean, Annie Lauzier, Caroline Normandin, Camille Lacarrière-Keita, Marie-France Bossanyi et Sonya Nassari qui ont tout d’abord participé activement à l’avancée de ce projet mais surtout avec qui j’ai passé d’excellents moments. Merci à vous de m’avoir supporté tous les jours pendant toutes ces années.

Je remercie l’ensemble des membres de l’équipe du Pr Francois-Michel Boisvert pour leurs aiguillages dans ce projet qui se sont souvent avérés d’une grande aide.

Enfin, un grand merci à ma mère, mon père et ma sœur pour leur soutien indéfectible. Merci à Revancha et à Caro pour tous ces bons moments et d’avoir permis à mon esprit de s’évader de temps en temps.

196 197

7LISTE DES RÉFÉRENCES

Aguilera-Romero, Auxiliadora, Joanna Kaminska, Anne Spang, Howard Riezman, and Manuel Muñiz. 2008. “The Yeast P24 Complex Is Required for the Formation of COPI Retrograde Transport Vesicles from the Golgi Apparatus.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200710025. Alanko, Jonna, Anja Mai, Guillaume Jacquemet, Kristine Schauer, Riina Kaukonen, Markku Saari, Bruno Goud, and Johanna Ivaska. 2015. “Integrin Endosomal Signalling Suppresses Anoikis.” Nature Cell Biology 17 (11): 1412–21. https://doi.org/10.1038/ncb3250. Alekhina, Olga, Ezra Burstein, and Daniel D. Billadeau. 2017. “Cellular Functions of WASP Family Proteins at a Glance.” Journal of Cell Science 130 (14): 2235–41. https://doi.org/10.1242/jcs.199570. Ali, Moazzam, Fai Leung, and C Field. 2014. “The Ancient Small GTPase Rab21 Functions in Intermediate Endocytic Steps in Trypanosomes” 13 (2): 304–19. https://doi.org/10.1128/EC.00269-13. Anantharaman, Vivek, and L Aravind. 2002. “The GOLD Domain, a Novel Protein Module Involved in Golgi Function and Secretion.” Genome Biology 3 (5): research0023. http://www.ncbi.nlm.nih.gov/pubmed/12049664. Antonny, Bruno, Christopher Burd, Pietro De Camilli, Elizabeth Chen, Oliver Daumke, Katja Faelber, Marijn Ford, et al. 2016. “Membrane Fission by Dynamin: What We Know and What We Need to Know.” The EMBO Journal. https://doi.org/10.15252/embj.201694613. Appenzeller-Herzog, Christian, and Hans-Peter Hauri. 2006. “The ER-Golgi Intermediate Compartment (ERGIC): In Search of Its Identity and Function.” Journal of Cell Science 119 (Pt 11): 2173–83. https://doi.org/10.1242/jcs.03019. Arakel, Eric C, and Blanche Schwappach. 2018. “Formation of COPI-Coated Vesicles at a Glance.” Journal of Cell Science 131 (5). https://doi.org/10.1242/jcs.209890. Armstrong, John. 2000. “How Do Rab Proteins Function in Membrane Traffic?” International Journal of Biochemistry and Cell Biology. https://doi.org/10.1016/S1357-2725(99)00112-0. Aviram, Naama, and Maya Schuldiner. 2017. “Targeting and Translocation of Proteins to the Endoplasmic Reticulum at a Glance.” Journal of Cell Science 130 (24): 4079–85. https://doi.org/10.1242/jcs.204396. Bahl, Kriti, Shuwei Xie, Gaelle Spagnol, Paul Sorgen, Naava Naslavsky, and Steve Caplan. 2016. “EHD3 Protein Is Required for Tubular Recycling Endosome Stabilization, and an Asparagine- Glutamic Acid Residue Pair within Its Eps15 Homology (EH) Domain Dictates Its Selective Binding to NPF Peptides.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M116.716407. Barr, Francis, and David G Lambright. 2010. “Rab GEFs and GAPs.” Current Opinion in Cell Biology 22 (4): 461–70. https://doi.org/10.1016/j.ceb.2010.04.007. Beckers, Con J.M., Marc R. Block, Benjamin S. Glick, James E. Rothman, and William E. Balch. 1989. “Vesicular Transport between the Endoplasmic Reticulum and the Golgi Stack Requires the NEM-Sensitive Fusion Protein.” Nature. https://doi.org/10.1038/339397a0. Belden, W J, and C Barlowe. 2001. “Distinct Roles for the Cytoplasmic Tail Sequences of Emp24p and Erv25p in Transport between the Endoplasmic Reticulum and Golgi Complex.” The Journal of Biological Chemistry 276 (46): 43040–48. https://doi.org/10.1074/jbc.M108113200. Belenkaya, Tatyana Y., Yihui Wu, Xiaofang Tang, Bo Zhou, Longqiu Cheng, Yagya V. Sharma, Dong Yan, Erica M. Selva, and Xinhua Lin. 2008. “The Retromer Complex Influences Wnt Secretion by Recycling Wntless from Endosomes to the Trans-Golgi Network.” Developmental Cell. https://doi.org/10.1016/j.devcel.2007.12.003. Béthune, Julien, Matthijs Kol, Julia Hoffmann, Inge Reckmann, Britta Brügger, and Felix Wieland.

197 198

2006. “Coatomer, the Coat Protein of COPI Transport Vesicles, Discriminates Endoplasmic Reticulum Residents from P24 Proteins.” Molecular and Cellular Biology 26 (21): 8011–21. https://doi.org/10.1128/MCB.01055-06. Béthune, Julien, and Felix T. Wieland. 2018. “Assembly of COPI and COPII Vesicular Coat Proteins on Membranes.” Annual Review of Biophysics 47 (1): 63–83. https://doi.org/10.1146/annurev- biophys-070317-033259. Bhuin, Tanmay, and Jagat Kumar Roy. 2014. “Rab Proteins: The Key Regulators of Intracellular Vesicle Transport.” Experimental Cell Research 328 (1): 1–19. https://doi.org/10.1016/j.yexcr.2014.07.027. Bi, Xiping, Joseph D Mancias, and Jonathan Goldberg. 2007. “Insights into COPII Coat Nucleation from the Structure of Sec23.Sar1 Complexed with the Active Fragment of Sec31.” Developmental Cell 13 (5): 635–45. https://doi.org/10.1016/j.devcel.2007.10.006. Blazek, Alisa D., Brian J. Paleo, and Noah Weisleder. 2015. “Plasma Membrane Repair: A Central Process for Maintaining Cellular Homeostasis.” Physiology 30 (6): 438. https://doi.org/10.1152/PHYSIOL.00019.2015. Blum, Robert, and Alexandra Lepier. 2008. “The Luminal Domain of P23 (Tmp21) Plays a Critical Role in P23 Cell Surface Trafficking.” Traffic 9 (9): 1530–50. https://doi.org/10.1111/j.1600- 0854.2008.00784.x. Blum, Robert, Fatima Pfeiffer, Peter Feick, Wolfgang Nastainczyk, Bärbel Kohler, and Karl-herbert Schäfer. 1999. “Intracellular Localization and in Vivo Trafficking of P24A and P23” 548: 537– 48. Blümer, Julia, Juliana Rey, Leif Dehmelt, Tomáǎ Maze, Yao Wen Wu, Philippe Bastiaens, Roger S. Goody, and Aymelt Itzen. 2013. “RabGEFs Are a Major Determinant for Specific Rab Membrane Targeting.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201209113. Boncompain, Gaelle, and Franck Perez. 2013. “The Many Routes of Golgi-Dependent Trafficking.” Histochemistry and Cell Biology 140 (3): 251–60. https://doi.org/10.1007/s00418-013-1124-7. Bos, Johannes L., Holger Rehmann, and Alfred Wittinghofer. 2007. “GEFs and GAPs: Critical Elements in the Control of Small G Proteins.” Cell. https://doi.org/10.1016/j.cell.2007.05.018. Bourne, Henry R., David A. Sanders, and Frank McCormick. 1991. “The GTPase Superfamily: Conserved Structure and Molecular Mechanism.” Nature. https://doi.org/10.1038/349117a0. Bremser, Martina, Walter Nickel, Michael Schweikert, Mariella Ravazzola, Mylène Amherdt, Christine A Hughes, Thomas H Söllner, James E Rothman, and Felix T Wieland. 1999. “Coupling of Coat Assembly and Vesicle Budding to Packaging of Putative Cargo Receptors.” Cell 96 (4): 495–506. https://doi.org/10.1016/S0092-8674(00)80654-6. Brighouse, Andrew, Joel B. Dacks, and Mark C. Field. 2010. “Rab Protein Evolution and the History of the Eukaryotic Endomembrane System.” Cellular and Molecular Life Sciences 67 (20): 3449–65. https://doi.org/10.1007/s00018-010-0436-1. Brodin, Lennart, and Oleg Shupliakov. 2018. “Retromer in Synaptic Function and Pathology” 10 (October): 1–9. https://doi.org/10.3389/fnsyn.2018.00037. Bugarcic, Andrea, Yang Zhe, Markus C. Kerr, John Griffin, Brett M. Collins, and Rohan D. Teasdale. 2011. “Vps26A and Vps26B Subunits Define Distinct Retromer Complexes.” Traffic. https://doi.org/10.1111/j.1600-0854.2011.01284.x. Burgo, Andrea, Emmanuel Sotirakis, Marie-Christine Simmler, Agathe Verraes, Christophe Chamot, Jeremy C Simpson, Letizia Lanzetti, Véronique Proux-Gillardeaux, and Thierry Galli. 2009. “Role of Varp, a Rab21 Exchange Factor and TI-VAMP/VAMP7 Partner, in Neurite Growth.” EMBO Reports 10 (10): 1117–24. https://doi.org/10.1038/embor.2009.186. Bykov, Yury S, Miroslava Schaffer, Svetlana O Dodonova, Sahradha Albert, Jürgen M Plitzko, Wolfgang Baumeister, Benjamin D Engel, and John AG Briggs. 2017. “The Structure of the COPI Coat Determined within the Cell.” ELife 6 (November). https://doi.org/10.7554/eLife.32493. Campa, Carlo Cosimo, and Emilio Hirsch. 2017. “Rab11 and Phosphoinositides: A Synergy of Signal

198 199

Transducers in the Control of Vesicular Trafficking.” Advances in Biological Regulation. https://doi.org/10.1016/j.jbior.2016.09.002. Cavet, Megan E., Jinjiang Pang, Guoyong Yin, and Bradford C. Berk. 2008. “An Epidermal Growth Factor (EGF) -Dependent Interaction between GIT1 and Sorting Nexin 6 Promotes Degradation of the EGF Receptor.” FASEB Journal. https://doi.org/10.1096/fj.07-094086. Chavrier, P, and B Goud. 1999. “The Role of ARF and Rab GTPases in Membrane Transport.” Current Opinion in Cell Biology 11 (4): 466–75. https://doi.org/10.1016/S0955- 0674(99)80067-2. Cherepanova, Natalia, Shiteshu Shrimal, and Reid Gilmore. 2016. “N-Linked Glycosylation and Homeostasis of the Endoplasmic Reticulum.” Current Opinion in Cell Biology 41: 57–65. https://doi.org/10.1016/j.ceb.2016.03.021. Cherfils, Jacqueline, and Mahel Zeghouf. 2013. “Regulation of Small GTPases by GEFs, GAPs, and GDIs.” Physiological Reviews 93 (1): 269–309. https://doi.org/10.1152/physrev.00003.2012. Christoforidis, Savvas, Heidi M. McBride, Robert D. Burgoyne, and Marino Zerial. 1999. “The Rab5 Effector EEA1 Is a Core Component of Endosome Docking.” Nature. https://doi.org/10.1038/17618. Ciufo, L F, and A Boyd. 2000. “Identification of a Lumenal Sequence Specifying the Assembly of Emp24p into P24 Complexes in the Yeast Secretory Pathway.” The Journal of Biological Chemistry 275 (12): 8382–88. https://doi.org/10.1074/jbc.275.12.8382. Clague, Michael J, and Sylvie Urbé. 2020 “Data Mining for Traffic Information.” Traffic 21 (1): 162– 168. https://doi.org/10.1111/tra.12702. Clairfeuille, Thomas, Caroline Mas, Audrey S.M. Chan, Zhe Yang, Maria Tello-Lafoz, Mintu Chandra, Jocelyn Widagdo, et al. 2016. “A Molecular Code for Endosomal Recycling of Phosphorylated Cargos by the SNX27-Retromer Complex.” Nature Structural and Molecular Biology 23 (10): 921–32. https://doi.org/10.1038/nsmb.3290. Costanzo, Michael, Benjamin VanderSluis, Elizabeth N Koch, Anastasia Baryshnikova, Carles Pons, Guihong Tan, Wen Wang, et al. 2016. “A Global Genetic Interaction Network Maps a Wiring Diagram of Cellular Function.” Science (New York, N.Y.) 353 (6306). https://doi.org/10.1126/science.aaf1420. Cruz, Mariel Delgado, and Kyoungtae Kim. 2019. “Review The Inner Workings of Intracellular Heterotypic and Homotypic Membrane Fusion Mechanisms.” Journal of Biosciences 44 (4): 1– 14. https://doi.org/10.1007/s12038-019-9913-3. Cui, Yi, Julian M. Carosi, Zhe Yang, Nicholas Ariotti, Markus C. Kerr, Robert G. Parton, Timothy J. Sargeant, and Rohan D. Teasdale. 2019. “Retromer Has a Selective Function in Cargo Sorting via Endosome Transport Carriers.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201806153. D’Arcangelo, Jennifer G., Jonathan Crissman, Silvere Pagant, Alenka Čopič, Catherine F. Latham, Erik L. Snapp, and Elizabeth A. Miller. 2015. “Traffic of P24 Proteins and COPII Coat Composition Mutually Influence Membrane Scaffolding.” Current Biology. https://doi.org/10.1016/j.cub.2015.03.029. Deng, Zhi-Hui, Timothy S. Gomez, Douglas G. Osborne, Christine A. Phillips-Krawczak, Jin-San Zhang, and Daniel D. Billadeau. 2015. “Nuclear FAM21 Participates in NF-ΚB-Dependent Gene Regulation in Pancreatic Cancer Cells.” Journal of Cell Science 128 (2): 373. https://doi.org/10.1242/JCS.161513. Denzel, A., F. Otto, A. Girod, R. Pepperkok, R. Watson, I. Rosewell, J.J.M. Bergeron, R.C.E. Solarie, and M.J. Owen. 2000. “The P24 Family Member P23 Is Required for Early Embryonic Development.” Current Biology 10 (1): 55–58. https://doi.org/10.1016/S0960-9822(99)00266- 3. Derivery, Emmanuel, Emmanuèle Helfer, Véronique Henriot, and Alexis Gautreau. 2012. “Actin Polymerization Controls the Organization of WASH Domains at the Surface of Endosomes.” PLoS ONE. https://doi.org/10.1371/journal.pone.0039774.

199 200

Derivery, Emmanuel, Carla Sousa, Jérémie J. Gautier, Bérangère Lombard, Damarys Loew, and Alexis Gautreau. 2009. “The Arp2/3 Activator WASH Controls the Fission of Endosomes through a Large Multiprotein Complex.” Developmental Cell. https://doi.org/10.1016/j.devcel.2009.09.010. Dominguez, M, K Dejgaard, J Füllekrug, S Dahan, A Fazel, J P Paccaud, D Y Thomas, J J Bergeron, and T Nilsson. 1998. “Gp25L/Emp24/P24 Protein Family Members of the Cis-Golgi Network Bind Both COP I and II Coatomer.” The Journal of Cell Biology 140 (4): 751–65. https://doi.org/10.1083/jcb.140.4.751. Dong, Rui, Yasunori Saheki, Sharan Swarup, Louise Lucast, J. Wade Harper, and Pietro De Camilli. 2016. “Endosome-ER Contacts Control Actin Nucleation and Retromer Function through VAP- Dependent Regulation of PI4P.” Cell 166 (2): 408–23. https://doi.org/10.1016/j.cell.2016.06.037. Donoso, Maribel, Jorge Cancino, Jiyeon Lee, Peter Van Kerkhof, Claudio Retamal, Guojun Bu, Alfonso Gonzalez, Alfredo Caceres, and Maria Paz Marzolo. 2009. “Polarized Traffic of LRP1 Involves AP1B and SNX17 Operating on Y-Dependent Sorting Motifs in Different Pathways.” Molecular Biology of the Cell. https://doi.org/10.1091/mbc.E08-08-0805. Driskell, Owen J., Aleksandr Mironov, Victoria J. Allan, and Philip G. Woodman. 2007. “Dynein Is Required for Receptor Sorting and the Morphogenesis of Early Endosomes.” Nature Cell Biology. https://doi.org/10.1038/ncb1525. Duleh, Steve N., and Matthew D. Welch. 2010. “WASH and the Arp2/3 Complex Regulate Endosome Shape and Trafficking.” Cytoskeleton. https://doi.org/10.1002/cm.20437. Dumas, John J., Zhongyuan Zhu, Joseph L. Connolly, and David G. Lambright. 1999. “Structural Basis of Activation and GTP Hydrolysis in Rab Proteins.” Structure. https://doi.org/10.1016/S0969-2126(99)80054-9. Egami, Youhei, and Nobukazu Araki. 2008. “Characterization of Rab21-Positive Tubular Endosomes Induced by PI3K Inhibitors.” Experimental Cell Research 314 (4): 729–37. https://doi.org/10.1016/j.yexcr.2007.11.018. Eisenberg-Bord, Michal, Nadav Shai, Maya Schuldiner, and Maria Bohnert. 2016. “A Tether Is a Tether Is a Tether: Tethering at Membrane Contact Sites.” Developmental Cell. https://doi.org/10.1016/j.devcel.2016.10.022. Elkin, Sarah R, Ashley M Lakoduk, and Sandra L Schmid. 2016. “Endocytic Pathways and Endosomal Trafficking: A Primer.” Wiener Medizinische Wochenschrift (1946) 166 (7–8): 196– 204. https://doi.org/10.1007/s10354-016-0432-7. Ellson, Chris D., Simon Andrews, Len R. Stephens, and Phil T. Hawkins. 2002. “The PX Domain: A New Phosphoinositide-Binding Module.” Journal of Cell Science. Emery, G., M. Rojo, and J. Gruenberg. 2000. “Coupled Transport of P24 Family Members.” Journal of Cell Science. Eyster, Craig A., Jason D. Higginson, Robert Huebner, Natalie Porat-Shliom, Roberto Weigert, Wells W. Wu, Rong Fong Shen, and Julie G. Donaldson. 2009. “Discovery of New Cargo Proteins That Enter Cells through Clathrin-Independent Endocytosis.” Traffic. https://doi.org/10.1111/j.1600-0854.2009.00894.x. Fischer, G., G. A. Mignery, M. Baumert, M. S. Perin, T. J. Hanson, P. M. Burger, R. Jahn, and T. C. Sudhof. 1990. “Rab3 Is a Small GTP-Binding Protein Exclusively Localized to Synaptic Vesicles.” Proceedings of the National Academy of Sciences of the United States of America. https://doi.org/10.1073/pnas.87.5.1988. Fjorback, Anja W., Matthew Seaman, Camilla Gustafsen, Arnela Mehmedbasic, Suzanne Gokool, Chengbiao Wu, Daniel Militz, et al. 2012. “Retromer Binds the FANSHY Sorting Motif in SorLA to Regulate Amyloid Precursor Protein Sorting and Processing.” Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.2272-11.2012. Flannagan, Ronald S., Valentin Jaumouillé, and Sergio Grinstein. 2012. “The Cell Biology of Phagocytosis.” Annual Review of Pathology: Mechanisms of Disease.

200 201

https://doi.org/10.1146/annurev-pathol-011811-132445. Frasa, Marieke A.M., Katja T. Koessmeier, M. Reza Ahmadian, and Vania M.M. Braga. 2012. “Illuminating the Functional and Structural Repertoire of Human TBC/RABGAPs.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm3267. Fujita, Morihisa, Reika Watanabe, Nina Jaensch, Maria Romanova-Michaelides, Tadashi Satoh, Masaki Kato, Howard Riezman, Yoshiki Yamaguchi, Yusuke Maeda, and Taroh Kinoshita. 2011. “Sorting of GPI-Anchored Proteins into ER Exit Sites by P24 Proteins Is Dependent on Remodeled GPI.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201012074. Füllekrug, J, T Suganuma, B L Tang, W Hong, B Storrie, and T Nilsson. 1999. “Localization and Recycling of Gp27 (Hp24gamma3): Complex Formation with Other P24 Family Members.” Molecular Biology of the Cell 10 (6): 1939–55. https://doi.org/10.1091/mbc.10.6.1939. G. Stupack, Dwayne, and Vicente A. Torres. 2012. “Rab5 in the Regulation of Cell Motility and Invasion.” Current Protein & Peptide Science. https://doi.org/10.2174/138920311795659461. Gabe Lee, Meng Tse, Ashwini Mishra, and David G. Lambright. 2009. “Structural Mechanisms for Regulation of Membrane Traffic by Rab GTPases.” Traffic. https://doi.org/10.1111/j.1600- 0854.2009.00942.x. Gallon, Matthew, Thomas Clairfeuille, Florian Steinberg, Caroline Mas, Rajesh Ghai, Richard B. Sessions, Rohan D. Teasdale, Brett M. Collins, and Peter J. Cullen. 2014. “A Unique PDZ Domain and Arrestin-like Fold Interaction Reveals Mechanistic Details of Endocytic Recycling by SNX27-Retromer.” Proceedings of the National Academy of Sciences of the United States of America. https://doi.org/10.1073/pnas.1410552111. Gallon, Matthew, and Peter J. Cullen. 2015. “Retromer and Sorting Nexins in Endosomal Sorting.” Biochemical Society Transactions 43 (1): 33–47. https://doi.org/10.1042/BST20140290. Ganley, Ian G., Kate Carroll, Lenka Bittova, and Suzanne Pfeffer. 2004. “Rab9 GTPase Regulates Late Endosome Size and Requires Effector Interaction for Its Stability.” Molecular Biology of the Cell. https://doi.org/10.1091/mbc.E04-08-0747. Gautreau, Alexis, Ksenia Oguievetskaia, and Christian Ungermann. 2014. “Function and Regulation of the Endosomal Fusion and Fission Machineries.” Cold Spring Harbor Perspectives in Biology. https://doi.org/10.1101/cshperspect.a016832. Ge, Jian, Qianxue Chen, Baohui Liu, Long Wang, Shenqi Zhang, and Baowei Ji. 2017. “Knockdown of Rab21 Inhibits Proliferation and Induces Apoptosis in Human Glioma Cells.” Cellular & Molecular Biology Letters 22 (1): 30. https://doi.org/10.1186/s11658-017-0062-0. Gillingham, Alison K., Rita Sinka, Isabel L. Torres, Kathryn S. Lilley, and Sean Munro. 2014. “Toward a Comprehensive Map of the Effectors of Rab GTPases.” Developmental Cell. https://doi.org/10.1016/j.devcel.2014.10.007. Gillingham, Alison K, Jessie Bertram, Farida Begum, and Sean Munro. 2019. “In Vivo Identification of GTPase Interactors by Mitochondrial Relocalization and Proximity Biotinylation,” 1–36. Glick, Benjamin S, and Alberto Luini. 2011. “Models for Golgi Traffic: A Critical Assessment.” Cold Spring Harbor Perspectives in Biology 3 (11): a005215. https://doi.org/10.1101/cshperspect.a005215. Goh, Lai Kuan, and Alexander Sorkin. 2013. “Endocytosis of Receptor Tyrosine Kinases.” Cold Spring Harbor Perspectives in Biology. https://doi.org/10.1101/cshperspect.a017459. Gomez, Timothy S., and Daniel D. Billadeau. 2009. “A FAM21-Containing WASH Complex Regulates Retromer-Dependent Sorting.” Developmental Cell 17 (5): 699–711. https://doi.org/10.1016/j.devcel.2009.09.009. Gomez, Timothy S., Jacquelyn A. Gorman, Amaia Artal Martinez de Narvajas, Alexander O. Koenig, and Daniel D. Billadeau. 2012. “Trafficking Defects in WASH-Knockout Fibroblasts Originate from Collapsed Endosomal and Lysosomal Networks.” Molecular Biology of the Cell. https://doi.org/10.1091/mbc.e12-02-0101. Gommel, D, L Orci, E M Emig, M J Hannah, M Ravazzola, W Nickel, J B Helms, F T Wieland, and K Sohn. 1999. “P24 and P23, the Major Transmembrane Proteins of COPI-Coated Transport

201 202

Vesicles, Form Hetero-Oligomeric Complexes and Cycle between the Organelles of the Early Secretory Pathway.” FEBS Letters 447 (2–3): 179–85. http://www.ncbi.nlm.nih.gov/pubmed/10214941. Gong, Ping, Jelita Roseman, Celia G Fernandez, Kulandaivelu S Vetrivel, Vytautas P Bindokas, Lois A Zitzow, Satyabrata Kar, Angèle T Parent, and Gopal Thinakaran. 2011. “Transgenic Neuronal Overexpression Reveals That Stringently Regulated P23 Expression Is Critical for Coordinated Movement in Mice.” Molecular Neurodegeneration 6 (1): 87. https://doi.org/10.1186/1750- 1326-6-87. Gonzalez, Lino, and Richard H. Scheller. 1999. “Regulation of Membrane Trafficking: Structural Insights from a Rab/Effector Complex.” Cell 96 (6): 755–58. https://doi.org/10.1016/S0092- 8674(00)80585-1. Guglielmo, Gianni M. Di, Christine Le Roy, Anne F. Goodfellow, and Jeffrey L. Wrana. 2003. “Distinct Endocytic Pathways Regulate TGF-β Receptor Signalling and Turnover.” Nature Cell Biology. https://doi.org/10.1038/ncb975. Guo, Yusong, Daniel W. Sirkis, and Randy Schekman. 2014. “Protein Sorting at the Trans -Golgi Network.” Annual Review of Cell and Developmental Biology 30 (1): 169–206. https://doi.org/10.1146/annurev-cellbio-100913-013012. Hansen, Carsten G., and Benjamin J. Nichols. 2009. “Molecular Mechanisms of Clathrin-Independent Endocytosis.” Journal of Cell Science. https://doi.org/10.1242/jcs.033951. Hao, Yi Heng, Jennifer M. Doyle, Saumya Ramanathan, Timothy S. Gomez, Da Jia, Ming Xu, Zhijian J. Chen, Daniel D. Billadeau, Michael K. Rosen, and Patrick Ryan Potts. 2013. “Regulation of WASH-Dependent Actin Polymerization and Protein Trafficking by Ubiquitination.” Cell 152 (5): 1051–64. https://doi.org/10.1016/j.cell.2013.01.051. Harbour, Michael E., Sophia Y.A. Breusegem, Robin Antrobus, Caroline Freeman, Evan Reid, and Matthew N.J. Seaman. 2010. “The Cargo-Selective Retromer Complex Is a Recruiting Hub for Protein Complexes That Regulate Endosomal Tubule Dynamics.” Journal of Cell Science. https://doi.org/10.1242/jcs.071472. Harrison, Megan S., Chia Sui Hung, Ting Ting Liu, Romain Christiano, Tobias C. Walther, and Christopher G. Burd. 2014. “A Mechanism for Retromer Endosomal Coat Complex Assembly with Cargo.” Proceedings of the National Academy of Sciences of the United States of America. https://doi.org/10.1073/pnas.1316482111. Harterink, Martin, Fillip Port, Magdalena J. Lorenowicz, Ian J. McGough, Marie Silhankova, Marco C. Betist, Jan R.T. Van Weering, et al. 2011. “A SNX3-Dependent Retromer Pathway Mediates Retrograde Transport of the Wnt Sorting Receptor Wntless and Is Required for Wnt Secretion.” Nature Cell Biology. https://doi.org/10.1038/ncb2281. Hartl, F. Ulrich, and Manajit Hayer-Hartl. 2009. “Converging Concepts of Protein Folding in Vitro and in Vivo.” Nature Structural and Molecular Biology 16 (6): 574–81. https://doi.org/10.1038/nsmb.1591. Hauri, H P, F Kappeler, H Andersson, and C Appenzeller. 2000. “ERGIC-53 and Traffic in the Secretory Pathway.” Journal of Cell Science 113 ( Pt 4) (February): 587–96. http://www.ncbi.nlm.nih.gov/pubmed/10652252. Hehnly, Heidi, and Mark Stamnes. 2007. “Regulating Cytoskeleton-Based Vesicle Motility.” FEBS Letters 581 (11): 2112–18. https://doi.org/10.1016/j.febslet.2007.01.094. Helenius, Ari, Ira Mellman, Doris Wall, and Ann Hubbard. 1983. “Endosomes.” Trends in Biochemical Sciences 8 (7): 245–50. https://doi.org/10.1016/0968-0004(83)90350-X. Helfer, Emmanuèle, Michael E. Harbour, Véronique Henriot, Goran Lakisic, Carla Sousa-Blin, Larisa Volceanov, Matthew N.J. Seaman, and Alexis Gautreau. 2013. “Endosomal Recruitment of the WASH Complex: Active Sequences and Mutations Impairing Interaction with the Retromer.” Biology of the Cell. https://doi.org/10.1111/boc.201200038. Henley, John R., Eugene W.A. Krueger, Barbara J. Oswald, and Mark A. McNiven. 1998. “Dynamin- Mediated Internalization of Caveolae.” Journal of Cell Biology.

202 203

https://doi.org/10.1083/jcb.141.1.85. Henne, William Mike, Emmanuel Boucrot, Michael Meinecke, Emma Evergren, Yvonne Vallis, Rohit Mittal, and Harvey T. McMahon. 2010. “FCHo Proteins Are Nucleators of Clathrin- Mediated Endocytosis.” Science. https://doi.org/10.1126/science.1188462. Hierro, Aitor, Adriana L. Rojas, Raul Rojas, Namita Murthy, Grégory Effantin, Andrey V. Kajava, Alasdair C. Steven, Juan S. Bonifacino, and James H. Hurley. 2007. “Functional Architecture of the Retromer Cargo-Recognition Complex.” Nature. https://doi.org/10.1038/nature06216. Hill, Michelle M., Michele Bastiani, Robert Luetterforst, Matthew Kirkham, Annika Kirkham, Susan J. Nixon, Piers Walser, et al. 2008. “PTRF-Cavin, a Conserved Cytoplasmic Protein Required for Caveola Formation and Function.” Cell. https://doi.org/10.1016/j.cell.2007.11.042. Hooper, S, C Gaggioli, and E Sahai. 2009. “A Chemical Biology Screen Reveals a Role for Rab21- Mediated Control of Actomyosin Contractility in Fibroblast-Driven Cancer Invasion.” British Journal of Cancer 102 (2): 392–402. https://doi.org/10.1038/sj.bjc.6605469. Howes, Mark T., Matthew Kirkham, James Riches, Katia Cortese, Piers J. Walser, Fiona Simpson, Michelle M. Hill, et al. 2010. “Clathrin-Independent Carriers Form a High Capacity Endocytic Sorting System at the Leading Edge of Migrating Cells.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201002119. Huotari, Jatta, and Ari Helenius. 2011. “Endosome Maturation.” EMBO Journal. https://doi.org/10.1038/emboj.2011.286. Hurst, Katie E, Kiley A Lawrence, Lety Reyes Angeles, Zhiwei Ye, Jie Zhang, Danyelle M Townsend, Nathan Dollo, and Jessica E Thaxton. n.d. “Endoplasmic Reticulum Protein Disulfide Isomerase Shapes T Cell E Ffi Cacy for Adoptive Cellular Therapy of Tumors” 4. Hutagalung, Alex H., and Peter J. Novick. 2011. “Role of Rab GTPases in Membrane Traffic and Cell Physiology.” Physiological Reviews 91 (1): 119–49. https://doi.org/10.1152/physrev.00059.2009. Hyvola, Noora, Aipo Diao, Eddie McKenzie, Alison Skippen, Shamshad Cockcroft, and Martin Lowe. 2006. “Membrane Targeting and Activation of the Lowe Syndrome Protein OCRL1 by Rab GTPases.” EMBO Journal. https://doi.org/10.1038/sj.emboj.7601274. Itzen, Aymelt, Olena Pylypenko, Roger S. Goody, Kirill Alexandrov, and Alexey Rak. 2006. “Nucleotide Exchange via Local Protein Unfolding - Structure of Rab8 in Complex with MSS4.” EMBO Journal. https://doi.org/10.1038/sj.emboj.7601044. Jackson, Antony P., Alexander Flett, Carl Smythe, Lindsay Hufton, Frank R. Wettey, and Elizabeth Smythe. 2003. “Clathrin Promotes Incorporation of Cargo into Coated Pits by Activation of the AP2 Adaptor Μ2 Kinase.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200304079. Jackson, Lauren P. 2014. “Structure and Mechanism of COPI Vesicle Biogenesis.” Current Opinion in Cell Biology 29 (August): 67–73. https://doi.org/10.1016/j.ceb.2014.04.009. Jackson, Lauren P, Michael Lewis, Helen M Kent, Melissa A Edeling, Philip R Evans, Rainer Duden, and David J Owen. 2012. “Molecular Basis for Recognition of Dilysine Trafficking Motifs by COPI.” Developmental Cell 23 (6): 1255–62. https://doi.org/10.1016/j.devcel.2012.10.017. Jean, S., S. Cox, S. Nassari, and A. A. Kiger. 2015. “Starvation-Induced MTMR13 and RAB21 Activity Regulates VAMP8 to Promote Autophagosome-Lysosome Fusion.” EMBO Reports 16 (3): 297–311. https://doi.org/10.15252/embr.201439464. Jean, Steve, Sarah Cox, Eric J Schmidt, Fred L Robinson, and Amy Kiger. 2012. “Sbf/MTMR13 Coordinates PI(3)P and Rab21 Regulation in Endocytic Control of Cellular Remodeling.” Molecular Biology of the Cell 23 (14): 2723–40. https://doi.org/10.1091/mbc.E12-05-0375. Jean, Steve, and Amy A Kiger. 2012. “Coordination between RAB GTPase Functions.” Nature Reviews. Molecular Cell Biology 13: 463. Jedd, Gregory, Celeste Richardson, Robert Litt, and Nava Segev. 1995. “The Ypt1 GTPase Is Essential for the First Two Steps of the Yeast Secretory Pathway.” Journal of Cell Biology. https://doi.org/10.1083/jcb.131.3.583. Jia, Da, Timothy S. Gomez, Daniel D. Billadeau, and Michael K. Rosen. 2012. “Multiple Repeat

203 204

Elements within the FAM21 Tail Link the WASH Actin Regulatory Complex to the Retromer.” Molecular Biology of the Cell. https://doi.org/10.1091/mbc.E11-12-1059. Jia, Da, Timothy S Gomez, Zoltan Metlagel, Junko Umetani, Zbyszek Otwinowski, Michael K Rosen, and Daniel D Billadeau. 2010. “WASH and WAVE Actin Regulators of the Wiskott-Aldrich Syndrome Protein (WASP) Family Are Controlled by Analogous Structurally Related Complexes.” Proceedings of the National Academy of Sciences of the United States of America 107 (23): 10442–47. https://doi.org/10.1073/pnas.0913293107. Jia, Rui, and Juan S Bonifacino. n.d. “Lysosome Positioning Influences MTORC2 and AKT Article Lysosome Positioning Influences MTORC2 and AKT Signaling.” Molecular Cell, 1–13. https://doi.org/10.1016/j.molcel.2019.05.009. Jimenez‐Orgaz, Ana, Arunas Kvainickas, Heike Nägele, Justin Denner, Stefan Eimer, Jörn Dengjel, and Florian Steinberg. 2018. “ Control of RAB 7 Activity and Localization through the Retromer‐TBC1D5 Complex Enables RAB 7‐dependent Mitophagy .” The EMBO Journal. https://doi.org/10.15252/embj.201797128. Jovic, Marko, Mahak Sharma, Juliati Rahajeng, and Steve Caplan. 2010. “The Early Endosome: A Busy Sorting Station for Proteins at the Crossroads.” Histology and Histopathology 25 (1): 99– 112. https://doi.org/10.14670/HH-25.99. Kaksonen, Marko, and Aurelien Roux. 2018. “Mechanisms of Clathrin-Mediated Endocytosis.” Nature Reviews Molecular Cell Biology 19 (5): 313–26. https://doi.org/10.1038/nrm.2017.132. Kanellos, Georgios, Jing Zhou, Hitesh Patel, Rachel A. Ridgway, David Huels, Christine B. Gurniak, Emma Sandilands, et al. 2015. “ADF and Cofilin1 Control Actin Stress Fibers, Nuclear Integrity, and Cell Survival.” Cell Reports. https://doi.org/10.1016/j.celrep.2015.10.056. Kauppi, Maria, Anne Simonsen, Bjørn Bremnes, Amandio Vieira, Judy Callaghan, Harald Stenmark, and Vesa M. Olkkonen. 2002. “The Small GTPase Rab22 Interacts with EEA1 and Controls Endosomal Membrane Trafficking.” Journal of Cell Science. Kazazic, Maja, Vibeke Bertelsen, Ketil Winther Pedersen, Tram Thu Vuong, Michael Vibo Grandal, Marianne Skeie Rødland, Linton M. Traub, Espen Stang, and Inger Helene Madshus. 2009. “Epsin 1 Is Involved in Recruitment of Ubiquitinated EGF Receptors into Clathrin-Coated Pits.” Traffic. https://doi.org/10.1111/j.1600-0854.2008.00858.x. Kerr, Markus C., Jennifer S. Bennetts, Fiona Simpson, Elaine C. Thomas, Cameron Flegg, Paul A. Gleeson, Carol Wicking, and Rohan D. Teasdale. 2005. “A Novel Mammalian Retromer Component, Vps26B.” Traffic. https://doi.org/10.1111/j.1600-0854.2005.00328.x. Khan, Amir R., and Julie Ménétrey. 2013. “Structural Biology of Arf and Rab GTPases’ Effector Recruitment and Specificity.” Structure. https://doi.org/10.1016/j.str.2013.06.016. Khurana, Taruna, Joseph A Brzostowski, and Alan R Kimmel. 2005. “A Rab21 / LIM-Only / CH- LIM Complex Regulates Mechanisms” 24 (13): 2254–64. https://doi.org/10.1038/sj.emboj.7600716. Kim, L. C., R. S. Cook, and J. Chen. 2017. “MTORC1 and MTORC2 in Cancer and the Tumor Microenvironment.” Oncogene. https://doi.org/10.1038/onc.2016.363. Kiral, Ferdi Ridvan, Friederike Elisabeth Kohrs, Eugene Jennifer Jin, and Peter Robin Hiesinger. 2018. “Rab GTPases and Membrane Trafficking in Neurodegeneration.” Current Biology 28 (8): R471–86. https://doi.org/10.1016/j.cub.2018.02.010. Koch, Daniel, Amrita Rai, Imtiaz Ali, Nathalie Bleimling, Timon Friese, Andreas Brockmeyer, Petra Janning, et al. 2016. “A Pull-down Procedure for the Identification of Unknown GEFs for Small GTPases.” Small GTPases. https://doi.org/10.1080/21541248.2016.1156803. Kollmar, Martin, Dawid Lbik, and Stefanie Enge. 2012. “Evolution of the Eukaryotic ARP2/3 Activators of the WASP Family: WASP, WAVE, WASH, and WHAMM, and the Proposed New Family Members WAWH and WAML.” BMC Research Notes. https://doi.org/10.1186/1756-0500-5-88. Korolchuk, Viktor I., Shinji Saiki, Maike Lichtenberg, Farah H. Siddiqi, Esteban A. Roberts, Sara Imarisio, Luca Jahreiss, et al. 2011. “Lysosomal Positioning Coordinates Cellular Nutrient

204 205

Responses.” Nature Cell Biology. https://doi.org/10.1038/ncb2204. Kovacs, Eva M., Suzie Verma, Radiya G. Ali, Aparna Ratheesh, Nicholas A. Hamilton, Anna Akhmanova, and Alpha S. Yap. 2011. “N-WASP Regulates the Epithelial Junctional Actin Cytoskeleton through a Non-Canonical Post-Nucleation Pathway.” Nature Cell Biology. https://doi.org/10.1038/ncb2290. Kovtun, Oleksiy, Natalya Leneva, Yury S. Bykov, Nicholas Ariotti, Rohan D. Teasdale, Miroslava Schaffer, Benjamin D. Engel, David J. Owen, John A.G. Briggs, and Brett M. Collins. 2018. “Structure of the Membrane-Assembled Retromer Coat Determined by Cryo-Electron Tomography.” Nature. https://doi.org/10.1038/s41586-018-0526-z. Kumari, Sudha, and Satyajit Mayor. 2008. “ARF1 Is Directly Involved in Dynamin-Independent Endocytosis.” Nature Cell Biology. https://doi.org/10.1038/ncb1666. Kurokawa, Kazuo, Yasuyuki Suda, and Akihiko Nakano. 2016. “Sar1 Localizes at the Rims of COPII-Coated Membranes in Vivo.” Journal of Cell Science 129 (17): 3231–37. https://doi.org/10.1242/jcs.189423. Kurten, Richard C., Anthony D. Eddington, Parag Chowdhury, Richard D. Smith, April D. Davidson, and Brian B. Shank. 2001. “Self-Assembly and Binding of a Sorting Nexin to Sorting Endosomes.” Journal of Cell Science. Kvainickas, Arunas, Ana Jimenez-Orgaz, Heike Nägele, Zehan Hu, Jörn Dengjel, and Florian Steinberg. 2017. “Cargo-Selective SNX-BAR Proteins Mediate Retromer Trimer Independent Retrograde Transport.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201702137. Kvainickas, Arunas, Heike Nägele, Wenjing Qi, Ladislav Dokládal, Ana Jimenez-Orgaz, Luca Stehl, Dipak Gangurde, et al. 2019. “Retromer and TBC1D5 Maintain Late Endosomal RAB7 Domains to Enable Amino Acid-Induced MTORC1 Signaling.” The Journal of Cell Biology 218 (9): 3019–38. https://doi.org/10.1083/jcb.201812110. Kvainickas, Arunas, Ana Jimenez Orgaz, Heike Nägele, Britta Diedrich, Kate J. Heesom, Jörn Dengjel, Peter J. Cullen, and Florian Steinberg. 2017. “Retromer- and WASH-Dependent Sorting of Nutrient Transporters Requires a Multivalent Interaction Network with ANKRD50.” Journal of Cell Science 130 (2): 382–95. https://doi.org/10.1242/jcs.196758. Lajoie, Patrick, and Ivan R. Nabi. 2010. “Lipid Rafts, Caveolae, and Their Endocytosis.” International Review of Cell and Molecular Biology. https://doi.org/10.1016/S1937- 6448(10)82003-9. Lam, Stephanie S., Jeffrey D. Martell, Kimberli J. Kamer, Thomas J. Deerinck, Mark H. Ellisman, Vamsi K. Mootha, and Alice Y. Ting. 2014. “Directed Evolution of APEX2 for Electron Microscopy and Proximity Labeling.” Nature Methods. https://doi.org/10.1038/nmeth.3179. Lamaze, Christophe, Annick Dujeancourt, Takeshi Baba, Charles G. Lo, Alexandre Benmerah, and Alice Dautry-Varsat. 2001. “Interleukin 2 Receptors and Detergent-Resistant Membrane Domains Define a Clathrin-Independent Endocytic Pathway.” Molecular Cell. https://doi.org/10.1016/S1097-2765(01)00212-X. Lamber, Ekaterina P., Ann Christin Siedenburg, and Francis A. Barr. 2019. “Rab Regulation by GEFs and GAPs during Membrane Traffic.” Current Opinion in Cell Biology 59: 34–39. https://doi.org/10.1016/j.ceb.2019.03.004. Langhans, M., T. Meckel, A. Krss, A. Lerich, and D. G. Robinson. 2012. “ERES (ER Exit Sites) and the ‘Secretory Unit Concept.’” Journal of Microscopy 247 (1): 48–59. https://doi.org/10.1111/j.1365-2818.2011.03597.x. Langhans, Markus, María Jesús Marcote, Peter Pimpl, Goretti Virgili-López, David G Robinson, and Fernando Aniento. 2008. “In Vivo Trafficking and Localization of P24 Proteins in Plant Cells.” Traffic 9 (5): 770–85. https://doi.org/10.1111/j.1600-0854.2008.00719.x. Lauzier, Annie, Josiann Normandeau-Guimond, Vanessa Vaillancourt-Lavigueur, Vincent Boivin, Martine Charbonneau, Nathalie Rivard, Michelle S. Scott, Claire M. Dubois, and Steve Jean. 2019. “Colorectal cancer cells respond differentially to autophagy inhibition in vivo.” Scientific Reports 9 (1): 11316. https://www.ncbi.nlm.nih.gov/pubmed/31383875

205 206

Lee, Marcus C.S., Elizabeth A. Miller, Jonathan Goldberg, Lelio Orci, and Randy Schekman. 2004. “BI-DIRECTIONAL PROTEIN TRANSPORT BETWEEN THE ER AND GOLGI.” Annual Review of Cell and Developmental Biology. https://doi.org/10.1146/annurev.cellbio.20.010403.105307. Li, Hewang, Hui-Fang Li, Robin A. Felder, Ammasi Periasamy, and Pedro A. Jose. 2008. “Rab4 and Rab11 Coordinately Regulate the Recycling of Angiotensin II Type I Receptor as Demonstrated by Fluorescence Resonance Energy Transfer Microscopy.” Journal of Biomedical Optics. https://doi.org/10.1117/1.2943286. Li, Jian, Peter J. Peters, Ming Bai, Jun Dai, Erik Bos, Tomas Kirchhausen, Konstantin V. Kandror, and Victor W. Hsu. 2007. “An ACAP1-Containing Clathrin Coat Complex for Endocytic Recycling.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200608033. Li, Ping, Yong hong Wu, Yan ting Zhu, Man xiang Li, and Hong hong Pei. 2019. “Requirement of Rab21 in LPS-Induced TLR4 Signaling and pro-Inflammatory Responses in Macrophages and Monocytes.” Biochemical and Biophysical Research Communications 508 (1): 169–76. https://doi.org/10.1016/j.bbrc.2018.11.074. Li, Qian, Yi Liu, and Miao Sun. 2017. “Autophagy and Alzheimer’s Disease.” Cellular and Molecular Neurobiology 37 (3): 377–88. https://doi.org/10.1007/s10571-016-0386-8. Ligeti, Erzsébet, Stefan Welti, and Klaus Scheffzek. 2012. “Inhibition and Termination of Physiological Responses by GTPase Activating Proteins.” Physiological Reviews. https://doi.org/10.1152/physrev.00045.2010. Lim, Jet Phey, and Paul A. Gleeson. 2011. “Macropinocytosis: An Endocytic Pathway for Internalising Large Gulps.” Immunology and Cell Biology. https://doi.org/10.1038/icb.2011.20. Liu, Chun-Chun, Yun-Na Zhang, Zhao-Yao Li, Jin-Xiu Hou, Jing Zhou, Lin Kan, Bin Zhou, and Pu- Yan Chen. 2017. “Rab5 and Rab11 Are Required for Clathrin-Dependent Endocytosis of Japanese Encephalitis Virus in BHK-21 Cells.” Journal of Virology. https://doi.org/10.1128/jvi.01113-17. Liu, Raymond, Maria Teresa Abreu-Blanco, Kevin C. Barry, Elena V. Linardopoulou, Gregory E. Osborn, and Susan M. Parkhurst. 2009. “Wash Functions Downstream of Rho and Links Linear and Branched Actin Nucleation Factors.” Development. https://doi.org/10.1242/dev.035246. Liu, Rong, Xiaoyong Zhi, and Qing Zhong. 2015. “ATG14 Controls SNARE-Mediated Autophagosome Fusion with a Lysosome.” Autophagy 11 (5): 847–49. https://doi.org/10.1080/15548627.2015.1037549. Liu, Shengchun, Kelley Bromley-Brits, Kun Xia, Jill Mittelholtz, Ruitao Wang, and Weihong Song. 2008. “TMP21 Degradation Is Mediated by the Ubiquitin-Proteasome Pathway.” European Journal of Neuroscience 28 (10): 1980–88. https://doi.org/10.1111/j.1460-9568.2008.06497.x. Liu, Shengchun, Si Zhang, Kelley Bromley-Brits, Fang Cai, Weihui Zhou, Kun Xia, Jill Mittelholtz, and Weihong Song. 2011. “Transcriptional Regulation of TMP21 by NFAT.” Molecular Neurodegeneration 6 (1): 21. https://doi.org/10.1186/1750-1326-6-21. Lobingier, Braden T., and Mark von Zastrow. 2019. “When Trafficking and Signaling Mix: How Subcellular Location Shapes G Protein-Coupled Receptor Activation of Heterotrimeric G Proteins.” Traffic 20 (2): 130–36. https://doi.org/10.1111/tra.12634. Lodish, Harvey, Arnold Berk, S Lawrence Zipursky, Paul Matsudaira, David Baltimore, and James Darnell. 2000. “Protein Sorting: Organelle Biogenesis and Protein Secretion.” https://www.ncbi.nlm.nih.gov/books/NBK21480/. Lucas, María, David C. Gershlick, Ander Vidaurrazaga, Adriana L. Rojas, Juan S. Bonifacino, and Aitor Hierro. 2016. “Structural Mechanism for Cargo Recognition by the Retromer Complex.” Cell 167 (6): 1623-1635.e14. https://doi.org/10.1016/j.cell.2016.10.056. Luo, Weibo, Yingfei Wang, and Georg Reiser. 2011. “Proteinase-Activated Receptors, Nucleotide P2Y Receptors, and μ-Opioid Receptor-1B Are under the Control of the Type I Transmembrane Proteins P23 and P24A in Post-Golgi Trafficking.” Journal of Neurochemistry 117 (1): 71–81. https://doi.org/10.1111/j.1471-4159.2011.07173.x.

206 207

Luzio, J. Paul, Brian A. Rous, Nicholas A. Bright, Paul R. Pryor, Barbara M. Mullock, and Robert C. Piper. 2000. “Lysosome-Endosome Fusion and Lysosome Biogenesis.” Journal of Cell Science. MacDonald, Ewan, Louise Brown, Arnaud Selvais, Han Liu, Thomas Waring, Daniel Newman, Jessica Bithell, et al. 2018. “HRS-WASH Axis Governs Actin-Mediated Endosomal Recycling and Cell Invasion.” The Journal of Cell Biology 217 (7): 2549–64. https://doi.org/10.1083/jcb.201710051. Majoul, I, M Straub, S W Hell, R Duden, and H D Söling. 2001. “KDEL-Cargo Regulates Interactions between Proteins Involved in COPI Vesicle Traffic: Measurements in Living Cells Using FRET.” Developmental Cell 1 (1): 139–53. http://www.ncbi.nlm.nih.gov/pubmed/11703931. Majoul, Irina, Martin Straub, Stefan W. Hell, Rainer Duden, and Hans-Dieter Soeling. 2001. “KDEL- Cargo Regulates Interactions between Proteins Involved in COPI Vesicle Traffic.” Developmental Cell 1 (1): 139–53. https://doi.org/10.1016/S1534-5807(01)00004-1. Mancias, Joseph D, and Jonathan Goldberg. 2008. “Structural Basis of Cargo Membrane Protein Discrimination by the Human COPII Coat Machinery.” The EMBO Journal 27 (21): 2918–28. https://doi.org/10.1038/emboj.2008.208. Marchesin, Valentina, Antonio Castro-Castro, Catalina Lodillinsky, Alessia Castagnino, Joanna Cyrta, Hélène Bonsang-Kitzis, Laetitia Fuhrmann, et al. 2015. “ARF6-JIP3/4 Regulate Endosomal Tubules for MT1-MMP Exocytosis in Cancer Invasion.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201506002. Marelli, Marcello, Jennifer J Smith, Sunhee Jung, Eugene Yi, Alexey I Nesvizhskii, Rowan H Christmas, Ramsey A Saleem, et al. 2004. “Quantitative Mass Spectrometry Reveals a Role for the GTPase Rho1p in Actin Organization on the Peroxisome Membrane.” The Journal of Cell Biology 167 (6): 1099–1112. https://doi.org/10.1083/jcb.200404119. Martinez-Outschoorn, Ubaldo E., Federica Sotgia, and Michael P. Lisanti. 2015. “Caveolae and Signalling in Cancer.” Nature Reviews Cancer. https://doi.org/10.1038/nrc3915. Martinez, Olivier, and Bruno Goud. 1998. “Rab Proteins.” Biochimica et Biophysica Acta 1404 (1– 2): 101–12. https://doi.org/10.1016/s0167-4889(98)00050-0. Martinez, Olivier, Anne Schmidt, Jean Salaméro, Bernard Hoflack, Michèle Roa, and Bruno Goud. 1994. “The Small GTP-Binding Protein Rab6 Functions in Intra-Golgi Transport.” Journal of Cell Biology. https://doi.org/10.1083/jcb.127.6.1575. Marton, M J, C R Vazquez de Aldana, H Qiu, K Chakraburtty, and A G Hinnebusch. 1997. “Evidence That GCN1 and GCN20, Translational Regulators of GCN4, Function on Elongating Ribosomes in Activation of EIF2alpha Kinase GCN2.” Molecular and Cellular Biology. https://doi.org/10.1128/mcb.17.8.4474. Mathieu, Cécile, Ines Li De La Sierra-Gallay, Romain Duval, Ximing Xu, Angélique Cocaign, Thibaut Léger, Gary Woffendin, et al. 2016. “Insights into Brain Glycogen Metabolism the Structure of Human Brain Glycogen Phosphorylase.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M116.738898. Matsuoka, K, L Orci, M Amherdt, S Y Bednarek, S Hamamoto, R Schekman, and T Yeung. 1998. “COPII-Coated Vesicle Formation Reconstituted with Purified Coat Proteins and Chemically Defined Liposomes.” Cell 93 (2): 263–75. https://doi.org/10.1016/s0092-8674(00)81577-9. Maxfield, Frederick R., and Timothy E. McGraw. 2004. “Endocytic Recycling.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm1315. Mayor, Satyajit, Robert G. Parton, and Julie G. Donaldson. 2014a. “Clathrin-Independent Pathways of Endocytosis.” Cold Spring Harbor Perspectives in Biology. https://doi.org/10.1101/cshperspect.a016758. Mayran, Nathalie, Robert G. Parton, and Jean Gruenberg. 2003. “Annexin II Regulates Multivesicular Endosome Biogenesis in the Degradation Pathway of Animal Cells.” EMBO Journal. https://doi.org/10.1093/emboj/cdg321. McBride, Heidi M., Vladimir Rybin, Carol Murphy, Angelika Giner, Rohan Teasdale, and Marino Zerial. 1999. “Oligomeric Complexes Link Rab5 Effectors with NSF and Drive Membrane

207 208

Fusion via Interactions between EEA1 and Syntaxin 13.” Cell. https://doi.org/10.1016/S0092- 8674(00)81966-2. McCaffrey, Kathleen, and Ineke Braakman. 2016. “Protein Quality Control at the Endoplasmic Reticulum.” Edited by Patricija van Oosten-Hawle. Essays in Biochemistry 60 (2): 227–35. https://doi.org/10.1042/EBC20160003. McCaughey, Janine, and David J. Stephens. 2018. “COPII-Dependent ER Export in Animal Cells: Adaptation and Control for Diverse Cargo.” Histochemistry and Cell Biology 150 (2): 119–31. https://doi.org/10.1007/s00418-018-1689-2. McGough, Ian J., Reinoud E.A. de Groot, Adam P. Jellett, Marco C. Betist, Katherine C. Varandas, Chris M. Danson, Kate J. Heesom, Hendrik C. Korswagen, and Peter J. Cullen. 2018. “SNX3- Retromer Requires an Evolutionary Conserved MON2:DOPEY2:ATP9A Complex to Mediate Wntless Sorting and Wnt Secretion.” Nature Communications. https://doi.org/10.1038/s41467- 018-06114-3. McMahon, Harvey T., and Emmanuel Boucrot. 2011. “Molecular Mechanism and Physiological Functions of Clathrin-Mediated Endocytosis.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm3151. McNally, Kerrie E., and Peter J. Cullen. 2018. “Endosomal Retrieval of Cargo: Retromer Is Not Alone.” Trends in Cell Biology 28 (10): 807–22. https://doi.org/10.1016/j.tcb.2018.06.005. McNally, Kerrie E., Rebecca Faulkner, Florian Steinberg, Matthew Gallon, Rajesh Ghai, David Pim, Paul Langton, et al. 2017. “Retriever Is a Multiprotein Complex for Retromer-Independent Endosomal Cargo Recycling.” Nature Cell Biology. https://doi.org/10.1038/ncb3610. Meer, Gerrit van, Dennis R Voelker, and Gerald W Feigenson. 2008. “Membrane Lipids: Where They Are and How They Behave.” Nature Reviews. Molecular Cell Biology 9 (2): 112–24. https://doi.org/10.1038/nrm2330. Mellman, Ira, and Graham Warren. 2000. “The Road Taken: Past and Future Foundations of Membrane Traffic.” Cell 100 (1): 99–112. https://doi.org/10.1016/S0092-8674(00)81687-6. Merithew, Eric, Scott Hatherly, John J. Dumas, Deirdre C. Lawe, Robin Heller-Harrison, and David G. Lambright. 2001. “Structural Plasticity of an Invariant Hydrophobic Triad in the Switch Regions of Rab GTPases Is a Determinant of Effector Recognition.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M009771200. Miserey-Lenkei, Stéphanie, Hugo Bousquet, Olena Pylypenko, Sabine Bardin, Ariane Dimitrov, Gaëlle Bressanelli, Raja Bonifay, et al. 2017. “Coupling Fission and Exit of RAB6 Vesicles at Golgi Hotspots through Kinesin-Myosin Interactions.” Nature Communications. https://doi.org/10.1038/s41467-017-01266-0. Mitrovic, Sandra, Houchaima Ben-Tekaya, Eva Koegler, Jean Gruenberg, and Hans Peter Hauri. 2008. “The Cargo Receptors Surf4, Endoplasmic Reticulum-Golgi Intermediate Compartment (ERGIC)-53, and P25 Are Required to Maintain the Architecture of ERGIC and Golgi.” Molecular Biology of the Cell. https://doi.org/10.1091/mbc.E07-10-0989. Monetta, Pablo, Ileana Slavin, Nahuel Romero, and Cecilia Alvarez. 2007. “Rab1b Interacts with GBF1 and Modulates Both ARF1 Dynamics and COPI Association.” Molecular Biology of the Cell 18 (7): 2400–2410. https://doi.org/10.1091/mbc.e06-11-1005. Montesinos, Juan Carlos, Noelia Pastor-Cantizano, David G Robinson, María Jesús Marcote, and Fernando Aniento. 2014. “Arabidopsis P24δ5 and P24δ9 Facilitate Coat Protein I-Dependent Transport of the K/HDEL Receptor ERD2 from the Golgi to the Endoplasmic Reticulum.” The Plant Journal : For Cell and Molecular Biology 80 (6): 1014–30. https://doi.org/10.1111/tpj.12700. Mori, Yasunori and Mitsunori Fukuda. 2013. “Rabex-5 determines the neurite localization of its downstream Rab proteins in hippocampal neurons.” Communicative & Integrative Biology 6 (5): e25433. https://doi.org/10.4161/cib.25433 Muñiz, M, C Nuoffer, H P Hauri, and H Riezman. 2000. “The Emp24 Complex Recruits a Specific Cargo Molecule into Endoplasmic Reticulum-Derived Vesicles.” The Journal of Cell Biology

208 209

148 (5): 925–30. https://doi.org/10.1083/jcb.148.5.925. Nagel, Benedikt M., Meike Bechtold, Luis Garcia Rodriguez, and Sven Bogdan. 2017. “Drosophila WASH Is Required for Integrin-Mediated Cell Adhesion, Cell Motility and Lysosomal Neutralization.” Journal of Cell Science 130 (2): 344–59. https://doi.org/10.1242/jcs.193086. Nakano, Akihiko, and Alberto Luini. 2010. “Passage through the Golgi.” Current Opinion in Cell Biology 22 (4): 471–78. https://doi.org/10.1016/J.CEB.2010.05.003. Nakano, Naoko, Yuki Tsuchiya, Kenro Kako, Kenryu Umezaki, Keigo Sano, Souichi Ikeno, Eri Otsuka, et al. 2017. “TMED10 Protein Interferes with Transforming Growth Factor (TGF)-β Signaling by Disrupting TGF-β Receptor Complex Formation.” Journal of Biological Chemistry 292 (10): 4099–4112. https://doi.org/10.1074/jbc.M116.769109. Naslavsky, Naava, and Steve Caplan. 2011. “EHD Proteins: Key Conductors of Endocytic Transport.” Trends in Cell Biology. https://doi.org/10.1016/j.tcb.2010.10.003. ———. 2018. “The Enigmatic Endosome – Sorting the Ins and Outs of Endocytic Trafficking.” https://doi.org/10.1242/jcs.216499. Niedergang, Florence, Stéphane Gasman, Nicolas Vitale, Claire Desnos, and Christophe Lamaze. 2017. “Meeting after Meeting: 20 Years of Discoveries by the Members of the Exocytosis- Endocytosis Club.” Biology of the Cell 109 (9): 339–53. https://doi.org/10.1111/boc.201700026. Norris, Anne, Prasad Tammineni, Simon Wang, Julianne Gerdes, Alexandra Murr, Kelvin Y. Kwan, Qian Cai, and Barth D. Grant. 2017. “SNX-1 and RME-8 Oppose the Assembly of HGRS-1/ ESCRT-0 Degradative Microdomains on Endosomes.” Proceedings of the National Academy of Sciences of the United States of America. https://doi.org/10.1073/pnas.1612730114. Opdam, Frank J M, Gera Kamps, Huib Croes, Hans Van Bokhoven, Leo A Ginsel, and Jack A M Fransen. 2000. “Expression of Rab Small GTPases in Epithelial Caco-2 Cells : Rab21 Is an Apically Located GTP-Binding Protein in Polarised Intestinal Epithelial Cells” 316: 308–16. Örtegren, Unn, Margareta Karlsson, Natascha Blazic, Maria Blomqvist, Fredrik H. Nystrom, Johanna Gustavsson, Pam Fredman, and Peter Strålfors. 2004. “Lipids and Glycosphingolipids in Caveolae and Surrounding Plasma Membrane of Primary Rat Adipocytes.” European Journal of Biochemistry. https://doi.org/10.1111/j.1432-1033.2004.04117.x. Osiecka-Iwan, Anna, Justyna Niderla-Bielinska, Anna Hyc, and Stanislaw Moskalewski. 2014. “Rat Chondrocyte-Associated Antigen Identified as Sialylated Transmembrane Protein Tmp21 Belonging to the P24 Protein Family.” Calcified Tissue International 94 (3): 348–52. https://doi.org/10.1007/s00223-013-9816-5. Pan, Xiaojing, Sudharshan Eathiraj, Mary Munson, and David G. Lambright. 2006. “TBC-Domain GAPs for Rab GTPases Accelerate GTP Hydrolysis by a Dual-Finger Mechanism.” Nature. https://doi.org/10.1038/nature04847. Parachoniak, Christine A., and Morag Park. 2012. “Dynamics of Receptor Trafficking in Tumorigenicity.” Trends in Cell Biology. https://doi.org/10.1016/j.tcb.2012.02.002. Pardossi-Piquard, Raphaëlle, Christopher Böhm, Fusheng Chen, Soshi Kanemoto, Frédéric Checler, Gerold Schmitt-Ulms, Peter St George-Hyslop, and Paul E. Fraser. 2009. “TMP21 Transmembrane Domain Regulates γ-Secretase Cleavage.” Journal of Biological Chemistry 284 (42): 28634–41. https://doi.org/10.1074/jbc.M109.059345. Park, Laura, Peter A. Thomason, Tobias Zech, Jason S. King, Douwe M. Veltman, Michael Carnell, Seiji Ura, Laura M. Machesky, and Robert H. Insall. 2013. “Cyclical Action of the WASH Complex: FAM21 and Capping Protein Drive WASH Recycling, Not Initial Recruitment.” Developmental Cell. https://doi.org/10.1016/j.devcel.2012.12.014. Parmar, Hirendrasinh B, and Roy Duncan. 2016. “A Novel Tribasic Golgi Export Signal Directs Cargo Protein Interaction with Activated Rab11 and AP-1-Dependent Golgi-Plasma Membrane Trafficking.” Molecular Biology of the Cell 27 (8): 1320–31. https://doi.org/10.1091/mbc.E15- 12-0845. Pastor-Cantizano, Noelia, Juan Carlos Montesinos, César Bernat-Silvestre, María Jesús Marcote, and

209 210

Fernando Aniento. 2016. “P24 Family Proteins: Key Players in the Regulation of Trafficking along the Secretory Pathway.” Protoplasma 253 (4): 967–85. https://doi.org/10.1007/s00709- 015-0858-6. Pavlos, Nathan J, and Peter A Friedman. 2017. “GPCR Signaling and Trafficking: The Long and Short of It The Classical View HHS Public Access.” Trends Endocrinol Metab 28 (3): 213–26. https://doi.org/10.1016/j.tem.2016.10.007. Pelkmans, Lucas, and Ari Helenius. 2002. “Endocytosis via Caveolae.” Traffic. https://doi.org/10.1034/j.1600-0854.2002.30501.x. Pellinen, Teijo, Antti Arjonen, Karoliina Vuoriluoto, Katja Kallio, Jack A M Fransen, and Johanna Ivaska. 2006. “Small GTPase Rab21 Regulates Cell Adhesion and Controls Endosomal Traffi c of β 1-Integrins” 173 (5): 767–80. https://doi.org/10.1083/jcb.200509019. Pellinen, Teijo, Saara Tuomi, Antti Arjonen, Maija Wolf, Henrik Edgren, Hannelore Meyer, Robert Grosse, et al. 2008. “Integrin Trafficking Regulated by Rab21 Is Necessary for Cytokinesis.” Developmental Cell 15 (3): 371–85. https://doi.org/10.1016/j.devcel.2008.08.001. Peotter, Jennifer, William Kasberg, Iryna Pustova, and Anjon Audhya. 2019. “COPII-Mediated Trafficking at the ER/ERGIC Interface.” Traffic 20 (7): 491–503. https://doi.org/10.1111/tra.12654. Peränen, Johan, Petri Auvinen, Hilkka Virta, Roger Wepf, and Kai Simons. 1996. “Rab8 Promotes Polarized Membrane Transport through Reorganization of Actin and Microtubules in Fibroblasts.” Journal of Cell Biology. https://doi.org/10.1083/jcb.135.1.153. Pereira-Leal, José B., and Miguel C. Seabra. 2001. “Evolution of the Rab Family of Small GTP- Binding Proteins.” Journal of Molecular Biology 313 (4): 889–901. https://doi.org/10.1006/jmbi.2001.5072. Pfeffer, Suzanne, and Dikran Aivazian. 2004. “TARGETING RAB GTPases TO DISTINCT MEMBRANE COMPARTMENTS” 5 (November): 886–96. https://doi.org/10.1038/nrm1500. Pfeffer, Suzanne R. 2001. “Rab GTPases: Specifying and Deciphering Organelle Identity and Function.” Trends in Cell Biology. https://doi.org/10.1016/S0962-8924(01)02147-X. ———. 2005. “Structural Clues to Rab GTPase Functional Diversity.” Journal of Biological Chemistry 280 (16): 15485–88. https://doi.org/10.1074/jbc.R500003200. Pfeffer, Suzanne R, and Doug Kellogg. 2017. “Rab GTPases : Master Regulators That Establish the Secretory and Endocytic Pathways.” https://doi.org/10.1091/mbc.E16-10-0737. Phillips-Krawczak, Christine A., Amika Singla, Petro Starokadomskyy, Zhihui Deng, Douglas G. Osborne, Haiying Li, Christopher J. Dick, et al. 2015. “COMMD1 Is Linked to the WASH Complex and Regulates Endosomal Trafficking of the Copper Transporter ATP7A.” Molecular Biology of the Cell 26 (1): 91. https://doi.org/10.1091/MBC.E14-06-1073. Piper, Robert C., and David J. Katzmann. 2007. “Biogenesis and Function of Multivesicular Bodies.” Annual Review of Cell and Developmental Biology. https://doi.org/10.1146/annurev.cellbio.23.090506.123319. Pizarro, Lorena, and Lorena Norambuena. 2014. “Regulation of Protein Trafficking: Posttranslational Mechanisms and the Unexplored Transcriptional Control.” Plant Science 225 (August): 24–33. https://doi.org/10.1016/j.plantsci.2014.05.004. Popoff, Vincent, Frank Adolf, Britta Brügger, and Felix Wieland. 2011. “COPI Budding within the Golgi Stack.” Cold Spring Harbor Perspectives in Biology 3 (11): a005231. https://doi.org/10.1101/cshperspect.a005231. Potelle, Sven, André Klein, and François Foulquier. 2015. “Golgi Post-Translational Modifications and Associated Diseases.” Journal of Inherited Metabolic Disease 38 (4): 741–51. https://doi.org/10.1007/s10545-015-9851-7. Presley, J F, N B Cole, T A Schroer, K Hirschberg, K J Zaal, and J Lippincott-Schwartz. 1997. “ER- to-Golgi Transport Visualized in Living Cells.” Nature 389 (6646): 81–85. https://doi.org/10.1038/38001. Presley, John F, Theresa H Ward, Andrea C Pfeifer, Eric D Siggia, Robert D Phair, and Jennifer

210 211

Lippincott-Schwartz. 2002. “Dissection of COPI and Arf1 Dynamics in Vivo and Role in Golgi Membrane Transport.” Nature 417 (6885): 187–93. https://doi.org/10.1038/417187a. Priya, Amulya, Inna V. Kalaidzidis, Yannis Kalaidzidis, David Lambright, and Sunando Datta. 2015. “Molecular Insights into Rab7-Mediated Endosomal Recruitment of Core Retromer: Deciphering the Role of Vps26 and Vps35.” Traffic. https://doi.org/10.1111/tra.12237. Priya, Amulya, Jini Sugatha, Sameena Parveen, Sandra Lacas-Gervais, Prateek Raj, Jérôme Gilleron, and Sunando Datta. 2017. “Essential and Selective Role of SNX12 in Transport of Endocytic and Retrograde Cargo.” Journal of Cell Science. https://doi.org/10.1242/jcs.201905. Puthenveedu, Manojkumar A., Benjamin Lauffer, Paul Temkin, Rachel Vistein, Peter Carlton, Kurt Thorn, Jack Taunton, Orion D. Weiner, Robert G. Parton, and Mark Von Zastrow. 2010. “Sequence-Dependent Sorting of Recycling Proteins by Actin-Stabilized Endosomal Microdomains.” Cell. https://doi.org/10.1016/j.cell.2010.10.003. Pylypenko, Olena, Hussein Hammich, I-mei Yu, and Anne Houdusse. 2018a. “Rab GTPases and Their Interacting Protein Partners : Structural Insights into Rab Functional Diversity” 9: 22–48. Pylypenko, Olena, Alexey Rak, Thomas Durek, Susanna Kushnir, Beatrice E. Dursina, Nicolas H. Thomae, Alexandru T. Constantinescu, et al. 2006. “Structure of Doubly Prenylated Ypt1:GDI Complex and the Mechanism of GDI-Mediated Rab Recycling.” EMBO Journal. https://doi.org/10.1038/sj.emboj.7600921. Quon, Evan, Yves Y. Sere, Neha Chauhan, Jesper Johansen, David P. Sullivan, Jeremy S. Dittman, William J. Rice, et al. 2018. “Endoplasmic Reticulum-Plasma Membrane Contact Sites Integrate Sterol and Phospholipid Regulation.” Edited by Sandra Schmid. PLOS Biology 16 (5): e2003864. https://doi.org/10.1371/journal.pbio.2003864. Radhakrishna, Harish, and Julie G. Donaldson. 1997. “ADP-Ribosylation Factor 6 Regulates a Novel Plasma Membrane Recycling Pathway.” Journal of Cell Biology. https://doi.org/10.1083/jcb.139.1.49. Razin, S. V. 2018. “Structural-Functional Organization of the Eukaryotic Cell Nucleus and Transcription Regulation: Introduction to This Special Issue of Biochemistry (Moscow).” Biochemistry (Moscow) 83 (4): 299–301. https://doi.org/10.1134/S0006297918040016. Rich, Peter R., and Amandine Maréchal. 2010. “The Mitochondrial Respiratory Chain.” Edited by Guy C. Brown and Michael P. Murphy. Essays in Biochemistry 47 (June): 1–23. https://doi.org/10.1042/bse0470001. Robinson, Fred L., and Jack E. Dixon. 2006. “Myotubularin Phosphatases: Policing 3- Phosphoinositides.” Trends in Cell Biology. https://doi.org/10.1016/j.tcb.2006.06.001. Rojas, Raul, Thijs van Vlijmen, Gonzalo A. Mardones, Yogikala Prabhu, Adriana L. Rojas, Shabaz Mohammed, Albert J.R. Heck, Graça Raposo, Peter van der Sluijs, and Juan S. Bonifacino. 2008. “Regulation of Retromer Recruitment to Endosomes by Sequential Action of Rab5 and Rab7.” The Journal of Cell Biology 183 (3): 513–26. https://doi.org/10.1083/jcb.200804048. Rosa-Ferreira, Cláudia, and Sean Munro. 2011. “Arl8 and SKIP Act Together to Link Lysosomes to Kinesin-1.” Developmental Cell. https://doi.org/10.1016/j.devcel.2011.10.007. Roth, T. F., and Porter K. R. 1964. “YOLK PROTEIN UPTAKE IN THE OOCYTE OF THE MOSQUITO AEDES AEGYPTI. L.” The Journal of Cell Biology. https://doi.org/10.1083/jcb.20.2.313. Rötter, Jutta, Roland P Kuiper, Gerrit Bouw, and Gerard J M Martens. 2002. “Cell-Type-Specific and Selectively Induced Expression of Members of the P24 Family of Putative Cargo Receptors.” Journal of Cell Science 115 (Pt 5): 1049–58. http://www.ncbi.nlm.nih.gov/pubmed/11870223. Rotty, Jeremy D., Congying Wu, and James E. Bear. 2013. “New Insights into the Regulation and Cellular Functions of the ARP2/3 Complex.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm3492. Roux, Anabel Lise Le, Xarxa Quiroga, Nikhil Walani, Marino Arroyo, and Pere Roca-Cusachs. 2019. “The Plasma Membrane as a Mechanochemical Transducer.” Philosophical Transactions of the Royal Society B: Biological Sciences 374 (1779). https://doi.org/10.1098/rstb.2018.0221.

211 212

Roux, Kyle J., Dae In Kim, Manfred Raida, and Brian Burke. 2012. “A Promiscuous Biotin Ligase Fusion Protein Identifies Proximal and Interacting Proteins in Mammalian Cells.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201112098. Rowland, Ashley A., Patrick J. Chitwood, Melissa J. Phillips, and Gia K. Voeltz. 2014. “ER Contact Sites Define the Position and Timing of Endosome Fission.” Cell. https://doi.org/10.1016/j.cell.2014.10.023. Saha, Sarbari, Debasna P. Panigrahi, Shankargouda Patil, and Sujit K. Bhutia. 2018. “Autophagy in Health and Disease: A Comprehensive Review.” Biomedicine & Pharmacotherapy 104 (August): 485–95. https://doi.org/10.1016/j.biopha.2018.05.007. Sallese, Michele, Monica Giannotta, and Alberto Luini. 2009. “Coordination of the Secretory Compartments via Inter-Organelle Signalling.” Seminars in Cell and Developmental Biology 20 (7): 801–9. https://doi.org/10.1016/j.semcdb.2009.04.004. Schoebel, Stefan, Lena Katharina Oesterlin, Wulf Blankenfeldt, Roger Sidney Goody, and Aymelt Itzen. 2009. “RabGDI Displacement by DrrA from Legionella Is a Consequence of Its Guanine Nucleotide Exchange Activity.” Molecular Cell. https://doi.org/10.1016/j.molcel.2009.11.014. Schwarz, Dianne S, and Michael D Blower. 2016. “The Endoplasmic Reticulum: Structure, Function and Response to Cellular Signaling.” Cellular and Molecular Life Sciences : CMLS 73 (1): 79– 94. https://doi.org/10.1007/s00018-015-2052-6. Seaman, Matthew N.J. 2004. “Cargo-Selective Endosomal Sorting for Retrieval to the Golgi Requires Retromer.” Journal of Cell Biology. https://doi.org/10.1083/jcb.200312034. ———. 2012. “The Retromer Complex-Endosomal Protein Recycling and Beyond.” Journal of Cell Science. https://doi.org/10.1242/jcs.103440. Seaman, Matthew N.J., Michael E. Harbour, Daniel Tattersall, Eliot Read, and Nicholas Bright. 2009. “Membrane Recruitment of the Cargo-Selective Retromer Subcomplex Is Catalysed by the Small GTPase Rab7 and Inhibited by the Rab-GAP TBC1D5.” Journal of Cell Science. https://doi.org/10.1242/jcs.048686. Seaman, Matthew Nj, and Caroline L Freeman. 2014. “Analysis of the Retromer Complex-WASH Complex Interaction Illuminates New Avenues to Explore in Parkinson Disease.” Communicative & Integrative Biology 7: e29483. https://doi.org/10.4161/cib.29483. Selkoe, Dennis J. 2001. “Alzheimer’s Disease: Genes, Proteins, and Therapy.” Physiological Reviews. https://doi.org/10.1152/physrev.2001.81.2.741. Shaughnessy, Ronan, and Arnaud Echard. 2018. “Rab35 GTPase and Cancer: Linking Membrane Trafficking to Tumorigenesis.” Traffic 19 (4): 247–52. https://doi.org/10.1111/tra.12546. Shih, Susan C., David J. Katzmann, Joshua D. Schnell, Myra Sutanto, Scott D. Emr, and Linda Hicke. 2002. “Epsins and Vps27p/Hrs Contain Ubiquitin-Binding Domains That Function in Receptor Endocytosis.” Nature Cell Biology. https://doi.org/10.1038/ncb790. Shin, Ji Hyun, So Jung Park, Doo Sin Jo, Na Yeon Park, Joon Bum Kim, Ji-Eun Bae, Yoon Kyung Jo, et al. 2019. “Down-Regulated TMED10 in Alzheimer Disease Induces Autophagy via ATG4B Activation.” Autophagy, March, 1–11. https://doi.org/10.1080/15548627.2019.1586249. Shin, John J.H., Alison K. Gillingham, Farida Begum, Jessica Chadwick, and Sean Munro. 2017. “TBC1D23 Is a Bridging Factor for Endosomal Vesicle Capture by Golgins at the Trans-Golgi.” Nature Cell Biology 19 (12): 1424–32. https://doi.org/10.1038/ncb3627. Short, Benjamin, Alexander Haas, and Francis A. Barr. 2005. “Golgins and GTPases, Giving Identity and Structure to the Golgi Apparatus.” Biochimica et Biophysica Acta - Molecular Cell Research. https://doi.org/10.1016/j.bbamcr.2005.02.001. Simonetti, Boris, and Peter J. Cullen. 2019. “Actin-Dependent Endosomal Receptor Recycling.” Current Opinion in Cell Biology 56: 22–33. https://doi.org/10.1016/j.ceb.2018.08.006. Simpson, J. C., Gareth Griffiths, Marianne Wessling-Resnick, Jack A M Fransen, Holly Bennett, and Arwyn T Jones. 2004. “A Role for the Small GTPase Rab21 in the Early Endocytic Pathway.” Journal of Cell Science 117 (26): 6297–6311. https://doi.org/10.1242/jcs.01560.

212 213

Singla, Amika, Alina Fedoseienko, Sai S.P. Giridharan, Brittany L. Overlee, Adam Lopez, Da Jia, Jie Song, et al. 2019. “Endosomal PI(3)P Regulation by the COMMD/CCDC22/CCDC93 (CCC) Complex Controls Membrane Protein Recycling.” Nature Communications. https://doi.org/10.1038/s41467-019-12221-6. Skalski, Michael, Qing Yi, Michelle J. Kean, Dennis W. Myers, Karla C. Williams, Angela Burtnik, and Marc G. Coppolino. 2010. “Lamellipodium Extension and Membrane Ruffling Require Different SNARE-Mediated Trafficking Pathways.” BMC Cell Biology. https://doi.org/10.1186/1471-2121-11-62. Sorkin, Alexander, and Mark Von Zastrow. 2009. “Endocytosis and Signalling: Intertwining Molecular Networks.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm2748. Spang, Anne. 2016. “Membrane Tethering Complexes in the Endosomal System.” Frontiers in Cell and Developmental Biology. https://doi.org/10.3389/fcell.2016.00035. Sprang, Stephen R. 1997. “G PROTEIN MECHANISMS:Insights from Structural Analysis.” Annual Review of Biochemistry 66 (1): 639–78. https://doi.org/10.1146/annurev.biochem.66.1.639. Stagg, Scott M., Cemal Gürkan, Douglas M. Fowler, Paul LaPointe, Ted R. Foss, Clinton S. Potter, Bridget Carragher, and William E. Balch. 2006. “Structure of the Sec13/31 COPII Coat Cage.” Nature 439 (7073): 234–38. https://doi.org/10.1038/nature04339. Stan, Radu V. 2005. “Structure of Caveolae.” Biochimica et Biophysica Acta - Molecular Cell Research. https://doi.org/10.1016/j.bbamcr.2005.08.008. Steinberg, Florian, Kate J. Heesom, Mark D. Bass, and Peter J. Cullen. 2012. “SNX17 Protects Integrins from Degradation by Sorting between Lysosomal and Recycling Pathways.” Journal of Cell Biology. https://doi.org/10.1083/jcb.201111121. Stenmark, Harald. 2009. “Rab GTPases as Coordinators of Vesicle Traffic.” Nature Reviews Molecular Cell Biology 10 (8): 513–25. https://doi.org/10.1038/nrm2728. Stenmark, Harald, Gaetano Vitale, Oliver Ullrich, and Marino Zerial. 1995. “Rabaptin-5 Is a Direct Effector of the Small GTPase Rab5 in Endocytic Membrane Fusion.” Cell. https://doi.org/10.1016/0092-8674(95)90120-5. Stephens, D J, and R Pepperkok. 2001. “Illuminating the Secretory Pathway: When Do We Need Vesicles?” Journal of Cell Science 114 (Pt 6): 1053–59. http://www.ncbi.nlm.nih.gov/pubmed/11228150. Stettler, Olivier, Ahmed Zahraoui, Kenneth L. Moya, and Bertrand Tavitian. 1992. “Expression of the Small GTP-Binding Protein Rab3A in the Adult Rat Brain.” Molecular and Cellular Neuroscience. https://doi.org/10.1016/1044-7431(92)90062-7. Strating, Jeroen R.P.M., and Gerard J.M. Martens. 2009. “The P24 Family and Selective Transport Processes at the ER-Golgi Interface.” Biology of the Cell 101 (9): 495–509. https://doi.org/10.1042/BC20080233. Strating, Jeroen R P M, Nick H M van Bakel, Jack A M Leunissen, and Gerard J M Martens. 2009. “A Comprehensive Overview of the Vertebrate P24 Family: Identification of a Novel Tissue- Specifically Expressed Member.” Molecular Biology and Evolution 26 (8): 1707–14. https://doi.org/10.1093/molbev/msp099. Strom, Molly, Alistair N. Hume, Abul K. Tarafder, Eleni Barkagianni, and Miguel C. Seabra. 2002. “A Family of Rab27-Binding Proteins: Melanophilin Links Rab27a and Myosin Va Function in Melanosome Transport.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M202574200. Südhof, Thomas C., and James E. Rothman. 2009. “Membrane Fusion: Grappling with SNARE and SM Proteins.” Science. https://doi.org/10.1126/science.1161748. Sun, Zhenzhen, Yujie Xie, Yintong Chen, Qinghu Yang, Zhenzhen Quan, Rongji Dai, and Hong Qing. 2017. “Rab21, a Novel PS1 Interactor, Regulates γ-Secretase Activity via PS1 Subcellular Distribution.” Molecular Neurobiology, May. https://doi.org/10.1007/s12035-017-0606-3. Taguchi, Tomohiko. 2013. “Emerging Roles of Recycling Endosomes.” Journal of Biochemistry.

213 214

https://doi.org/10.1093/jb/mvt034. Takahashi, Senye, Keiji Kubo, Satoshi Waguri, Atsuko Yabashi, Hye Won Shin, Yohei Katoh, and Kazuhisa Nakayama. 2012. “Rab11 Regulates Exocytosis of Recycling Vesicles at the Plasma Membrane.” Journal of Cell Science. https://doi.org/10.1242/jcs.102913. Takida, Satoshi, Yusuke Maeda, and Taroh Kinoshita. 2008. “Mammalian GPI-Anchored Proteins Require P24 Proteins for Their Efficient Transport from the ER to the Plasma Membrane.” The Biochemical Journal 409 (2): 555–62. https://doi.org/10.1042/BJ20070234. Temkin, Paul, Ben Lauffer, Stefanie Jäger, Peter Cimermancic, Nevan J. Krogan, and Mark Von Zastrow. 2011. “SNX27 Mediates Retromer Tubule Entry and Endosome-to-Plasma Membrane Trafficking of Signalling Receptors.” Nature Cell Biology. https://doi.org/10.1038/ncb2252. Theiler, Romina, Morihisa Fujita, Masamichi Nagae, Yoshiki Yamaguchi, Yusuke Maeda, and Taroh Kinoshita. 2014. “The α-Helical Region in P24γ2 Subunit of P24 Protein Cargo Receptor Is Pivotal for the Recognition and Transport of Glycosylphosphatidylinositol-Anchored Proteins.” The Journal of Biological Chemistry 289 (24): 16835–43. https://doi.org/10.1074/jbc.M114.568311. Thomas, Janice D., Yan Jie Zhang, Yue Hua Wei, Jun Hung Cho, Laura E. Morris, Hui Yun Wang, and X. F.Steven Zheng. 2014. “Rab1A Is an MTORC1 Activator and a Colorectal Oncogene.” Cancer Cell 26 (5): 754–69. https://doi.org/10.1016/j.ccell.2014.09.008. Tisdale, Ellen J., Jeffrey R. Bourne, Roya Khosravi-Far, Channing J. Der, and W. E. Balch. 1992. “GTP-Binding Mutants of Rab1 and Rab2 Are Potent Inhibitors of Vesicular Transport from the Endoplasmic Reticulum to the Golgi Complex.” Journal of Cell Biology. https://doi.org/10.1083/jcb.119.4.749. Touchot, N, P Chardin, and A Tavitian. 1987. “Four Additional Members of the Ras Gene Superfamily Isolated by an Oligonucleotide Strategy: Molecular Cloning of YPT-Related CDNAs from a Rat Brain Library.” Proceedings of the National Academy of Sciences of the United States of America 84 (23): 8210–14. https://doi.org/10.1073/pnas.84.23.8210. Ungewickell, Ernst J., and Lars Hinrichsen. 2007. “Endocytosis: Clathrin-Mediated Membrane Budding.” Current Opinion in Cell Biology 19 (4): 417–25. https://doi.org/10.1016/j.ceb.2007.05.003. Vanlandingham, Phillip A., and Brian P. Ceresa. 2009. “Rab7 Regulates Late Endocytic Trafficking Downstream of Multivesicular Body Biogenesis and Cargo Sequestration.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M809277200. Verboon, Jeffrey M., Hector Rincon-Arano, Timothy R. Werwie, Jeffrey J. Delrow, David Scalzo, Vivek Nandakumar, Mark Groudine, and Susan M. Parkhurst. 2015. “Wash Interacts with Lamin and Affects Global Nuclear Organization.” Current Biology 25 (6): 804–10. https://doi.org/10.1016/j.cub.2015.01.052. Vetrivel, Kulandaivelu S, Ping Gong, James W Bowen, Haipeng Cheng, Ying Chen, Meghan Carter, Phuong D Nguyen, et al. 2007. “Dual Roles of the Transmembrane Protein P23/TMP21 in the Modulation of Amyloid Precursor Protein Metabolism.” Molecular Neurodegeneration 2 (1): 4. https://doi.org/10.1186/1750-1326-2-4. Vetrivel, Kulandaivelu S, Anitha Kodam, Ping Gong, Ying Chen, Angèle T Parent, Satyabrata Kar, and Gopal Thinakaran. 2008. “Localization and Regional Distribution of P23/TMP21 in the Brain.” Neurobiology of Disease 32 (1): 37–49. https://doi.org/10.1016/j.nbd.2008.06.012. Vetter, I. R., and A. Wittinghofer. 2001. “The Guanine Nucleotide-Binding Switch in Three Dimensions.” Science. https://doi.org/10.1126/science.1062023. Vetter, Melanie, Ralf Stehle, Claire Basquin, and Esben Lorentzen. 2015. “Structure of Rab11-FIP3- Rabin8 Reveals Simultaneous Binding of FIP3 and Rabin8 Effectors to Rab11.” Nature Structural and Molecular Biology. https://doi.org/10.1038/nsmb.3065. Vicinanza, Mariella, Giovanni D’Angelo, Antonella Di Campli, and Maria Antonietta De Matteis. 2008. “Function and Dysfunction of the PI System in Membrane Trafficking.” EMBO Journal 27 (19): 2457–70. https://doi.org/10.1038/emboj.2008.169.

214 215

Vitavska, Olga, Hans Merzendorfer, and Helmut Wieczorek. 2005. “The V-ATPase Subunit C Binds to Polymeric F-Actin as Well as to Monomeric G-Actin and Induces Cross-Linking of Actin Filaments.” Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M406797200. Voeltz, Gia K., and Francis A. Barr. 2013. “Cell Organelles.” Current Opinion in Cell Biology 25 (4): 403–5. https://doi.org/10.1016/j.ceb.2013.06.001. Wandinger-Ness, Angela, and Marino Zerial. 2014. “Rab Proteins and the Compartmentalization of the Endosomal System.” Cold Spring Harbor Perspectives in Biology. https://doi.org/10.1101/cshperspect.a022616. Wang, HongBin, Liqing Xiao, and Marcelo G. Kazanietz. 2011. “P23/Tmp21 Associates with Protein Kinase Cδ (PKCδ) and Modulates Its Apoptotic Function.” Journal of Biological Chemistry 286 (18): 15821–31. https://doi.org/10.1074/jbc.M111.227991. Wang, Jing, Alina Fedoseienko, Baoyu Chen, Ezra Burstein, Da Jia, and Daniel D Billadeau. 2018. “Endosomal Receptor Trafficking : Retromer and Beyond,” no. April: 578–90. https://doi.org/10.1111/tra.12574. Wang, Na, Wenying Meng, Rongrong Jia, Shihao Xiang. 2019. “Rab GTPase 21 mediates caerulin- induced TRAF3-MKK3-p38 activation and acute pancreatitis response.” Biochemical and Biophysical Research Communications 518 (1): 50–58. https://doi.org/10.1016/j.bbrc.2019.08.007 Wang, Ting, Robert Grabski, Elizabeth Sztul, and Jesse C. Hay. 2015. “P115-SNARE Interactions: A Dynamic Cycle of P115 Binding Monomeric SNARE Motifs and Releasing Assembled Bundles.” Traffic. https://doi.org/10.1111/tra.12242. Wang, Tuanlao, Liangcheng Li, and Wanjin Hong. 2017. “SNARE Proteins in Membrane Trafficking.” Traffic 18 (12): 767–75. https://doi.org/10.1111/tra.12524. Wang, Yanzhuang, and Shijiao Huang. 2017. “Golgi Structure Formation, Function, and Post- Translational Modifications in Mammalian Cells.” F1000Research 6: 1–13. https://doi.org/10.12688/f1000research.11900.1. Wassmer, Thomas, Naomi Attar, Martin Harterink, Jan R.T. van Weering, Colin J. Traer, Jacqueline Oakley, Bruno Goud, et al. 2009. “The Retromer Coat Complex Coordinates Endosomal Sorting and Dynein-Mediated Transport, with Carrier Recognition by the Trans-Golgi Network.” Developmental Cell. https://doi.org/10.1016/j.devcel.2009.04.016. Weering, Jan R.T. Van, Richard B. Sessions, Colin J. Traer, Daniel P. Kloer, Vikram K. Bhatia, Dimitrios Stamou, Sven R. Carlsson, James H. Hurley, and Peter J. Cullen. 2012. “Molecular Basis for SNX-BAR-Mediated Assembly of Distinct Endosomal Sorting Tubules.” EMBO Journal. https://doi.org/10.1038/emboj.2012.283. Weering, Jan R.T. van, Paul Verkade, and Peter J. Cullen. 2012. “SNX-BAR-Mediated Endosome Tubulation Is Co-Ordinated with Endosome Maturation.” Traffic. https://doi.org/10.1111/j.1600-0854.2011.01297.x. Wei, Xiangyong, Lingfei Luo, and Jinzi Chen. 2019. “Roles of MTOR Signaling in Tissue Regeneration” 1. Weissman, Jacques T., H. Plutner, and William E. Balch. 2001. “The Mammalian Guanine Nucleotide Exchange Factor MSec12 Is Essential for Activation of the Sar1 GTPase Directing Endoplasmic Reticulum Export.” Traffic 2 (7): 465–75. https://doi.org/10.1034/j.1600- 0854.2001.20704.x. Wendeler, Markus W, Jean-Pierre Paccaud, and Hans-Peter Hauri. 2007. “Role of Sec24 Isoforms in Selective Export of Membrane Proteins from the Endoplasmic Reticulum.” EMBO Reports 8 (3): 258–64. https://doi.org/10.1038/sj.embor.7400893. Westrate, L.M., J.E. Lee, W.A. Prinz, and G.K. Voeltz. 2015. “Form Follows Function: The Importance of Endoplasmic Reticulum Shape.” Annual Review of Biochemistry 84 (1): 791– 811. https://doi.org/10.1146/annurev-biochem-072711-163501. Williams, Roger L., and Sylvie Urbé. 2007. “The Emerging Shape of the ESCRT Machinery.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm2162.

215 216

Xia, Pengyan, Shuo Wang, Ying Du, Zhenao Zhao, Lei Shi, Lei Sun, Guanling Huang, et al. 2013. “WASH Inhibits Autophagy through Suppression of Beclin 1 Ubiquitination.” EMBO Journal. https://doi.org/10.1038/emboj.2013.189. Xia, Pengyan, Shuo Wang, Guanling Huang, Ying Du, Pingping Zhu, Man Li, and Zusen Fan. 2014. “RNF2 Is Recruited by WASH to Ubiquitinate AMBRA1 Leading to Downregulation of Autophagy.” Cell Research. https://doi.org/10.1038/cr.2014.85. Xia, Pengyan, Shuo Wang, Guanling Huang, Pingping Zhu, Man Li, Buqing Ye, Ying Du, and Zusen Fan. 2014. “WASH Is Required for the Differentiation Commitment of Hematopoietic Stem Cells in a C-Myc-Dependent Manner.” Journal of Experimental Medicine. https://doi.org/10.1084/jem.20140169. Xing, Yi, Till Böcking, Matthias Wolf, Nikolaus Grigorieff, Tomas Kirchhausen, and Stephen C. Harrison. 2010. “Structure of Clathrin Coat with Bound Hsc70 and Auxilin: Mechanism of Hsc70-Facilitated Disassembly.” EMBO Journal. https://doi.org/10.1038/emboj.2009.383. Xu, Xiaobo, Hongqiang Gao, Jian Qin, Liu He, and Wenyong Liu. 2015. “TMP21 Modulates Cell Growth in Papillary Thyroid Cancer Cells by Inducing Autophagy through Activation of the AMPK / MTOR Pathway” 8 (9): 10824–31. Yang, Xi, Yanquan Zhang, Shan Li, Chunxiao Liu, Zhe Jin, Yinyin Wang, Fangli Ren, and Zhijie Chang. 2012. “Rab21 Attenuates EGF-Mediated MAPK Signaling through Enhancing EGFR Internalization and Degradation.” Biochemical and Biophysical Research Communications 421 (4): 651–57. https://doi.org/10.1016/j.bbrc.2012.04.049. Yuan, Qianying, Chunguang Ren, Wenwen Xu, Björn Petri, Jiasheng Zhang, Yong Zhang, Paul Kubes, Dianqing Wu, and Wenwen Tang. 2017. “PKN1 directs polarized RAB21 vesicle trafficking via phosphorylation of RPH3A and is important for neutrophil adhesion and ischemia-reperfusion injury.” Cell Reports 19(12):2586–2597. http://dx.doi.org/10.1016/j.celrep.2017.05.080 Zavodszky, Eszter, Matthew N.J. Seaman, Kevin Moreau, Maria Jimenez-Sanchez, Sophia Y. Breusegem, Michael E. Harbour, and David C. Rubinsztein. 2014. “Mutation in VPS35 Associated with Parkinson’s Disease Impairs WASH Complex Association and Inhibits Autophagy.” Nature Communications. https://doi.org/10.1038/ncomms4828. Zerial, Marino, and Heidi McBride. 2001. “Rab Proteins as Membrane Organizers.” Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/35052055. Zhang, Hong, and Junjie Hu. 2016. “Shaping the Endoplasmic Reticulum into a Social Network.” Trends in Cell Biology 26 (12): 934–43. https://doi.org/10.1016/j.tcb.2016.06.002. Zhang, Liling, and Allen Volchuk. 2010. “P24 Family Type 1 Transmembrane Proteins Are Required for Insulin Biosynthesis and Secretion in Pancreatic β-Cells.” FEBS Letters 584 (11): 2298– 2304. https://doi.org/10.1016/j.febslet.2010.03.041. Zhang, X., Xi He, Xin-Yuan Fu, and Zhijie Chang. 2006. “Varp Is a Rab21 Guanine Nucleotide Exchange Factor and Regulates Endosome Dynamics.” Journal of Cell Science 119 (6): 1053– 62. https://doi.org/10.1242/jcs.02810. Zhang, Yun-Na, Ya-Yun Liu, Fu-Chuan Xiao, Chun-Chun Liu, Xiao-Dong Liang, Jing Chen, Jing Zhou. 2018. “Rab5, Rab7, and Rab11 Are Required for Caveola-Dependent Endocytosis of Classical Swine Fever Virus in Porcine Alveolar Macrophages.” Journal of Virology. https://doi.org/10.1128/jvi.00797-18. Zhu, Guangyu, Jia Chen, Jay Liu, Joseph S Brunzelle, Bo Huang, Nancy Wakeham, Simon Terzyan,1 Xuemei Li,2 Zihe Rao,2 Guangpu Li,3 and Xuejun C Zhanga. 2007. “Structure of the APPL1 BAR-PH domain and characterization of its interaction with Rab5.” EMBO Journal 26 (14): 3484–3493. https://doi.org/10.1038/sj.emboj.7601771

216 217

8ANNEXES

Suivront les articles 1 et 2 sous leur format PDF publiés dans EMBO Reports et Biology Open ainsi que les autorisations des co-auteurs pour leur utilisation dans cette thèse.

217

Article

APEX2-mediated RAB proximity labeling identifies a role for RAB21 in clathrin-independent cargo sorting

Tomas Del Olmo1, Annie Lauzier1, Caroline Normandin1, Raphaëlle Larcher1, Mia Lecours1, Dominique Jean1, Louis Lessard1, Florian Steinberg2, François-Michel Boisvert1 & Steve Jean1,*

Abstract traffic and membrane compartment identity is the RAB GTPase family [5]. RAB GTPases are central modulators of membrane trafficking. They RAB GTPases form the largest family of small GTPases with are under the dynamic regulation of activating guanine exchange nearly 70 members in humans [6]. RABs are involved at every step factors (GEFs) and inactivating GTPase-activating proteins (GAPs). of vesicular trafficking where they regulate sorting, fission, trans- Once activated, RABs recruit a large spectrum of effectors to port, tethering, and fusion events [7]. RABs cycle between their control trafficking functions of eukaryotic cells. Multiple proteomic inactive (GDP-bound) and active (GTP-bound) states through the studies, using pull-down or yeast two-hybrid approaches, have action of guanine exchange factors (GEFs), and are conversely identified a number of RAB interactors. However, due to the in vitro converted from their GTP-bound state to their GDP state by GTPase- nature of these approaches and inherent limitations of each tech- activating proteins (GAPs) [8]. RABs are often targeted by multiple nique, a comprehensive definition of RAB interactors is still lacking. GEFs and GAPs [9,10]. The differential actions of these GEFs or By comparing quantitative affinity purifications of GFP:RAB21 with GAPs can potentially direct a RAB to alternate compartments or APEX2-mediated proximity labeling of RAB4a, RAB5a, RAB7a, and specify particular functions [11]. Once recruited and activated at RAB21, we find that APEX2 proximity labeling allows for the their target membranes, RABs bind effectors in order to mediate comprehensive identification of RAB regulators and interactors. various functions [12]. RAB effectors recruit multiple classes of Importantly, through biochemical and genetic approaches, we proteins ranging from adaptor coat proteins to myosins, dyneins, establish a novel link between RAB21 and the WASH and retromer and signaling kinases [12]. Moreover, RABs also interact with vari- complexes, with functional consequences on cargo sorting. Hence, ous cargos, tethering complexes and cargo sorting complexes [13]. APEX2-mediated proximity labeling of RAB neighboring proteins One such example is the role of RAB7 in the recruitment of the represents a new and efficient tool to define RAB functions. retromer complex at endosomes [13,14]. The retromer complex is an evolutionary conserved complex [15], which regulates endo- Keywords APEX2; clathrin-independent endocytosis; RAB GTPases; retromer; some-to-Golgi retrograde transport [16] as well as endosome-to- WASH complex plasma membrane trafficking [17]. The core retromer complex is Subject Categories Membrane & Intracellular Transport; Methods & constituted of two subcomplexes, namely the VPS and SNX Resources complexes. The VPS complex (or cargo sorting complex) comprises DOI 10.15252/embr.201847192 | Received 5 October 2018 | Revised 6 December VPS26, VPS29, and VPS35 [18] and interacts with various cargos 2018 | Accepted 10 December 2018 | Published online 3 January 2019 either directly or through binding with various SNXs and accessory EMBO Reports (2019) 20:e47192 proteins [17,19,20]. The SNX subcomplex interacts with endosomes through PtdIns(3)P [21] and also modulates cargo sorting [22,23] as well as membrane tubulation [24]. In addition, the retromer inter- Introduction acts with the WASH complex, through an interaction between VPS35 and the WASH complex subunit FAM21 [19,25,26]. The Membrane trafficking, the vesicular transport of cellular constitu- WASH complex generates F-actin at endosomes to mediate cargo ents, is fundamental to cellular and organismal homeostasis [1–3]. sorting and tubule scission [27–30]. It is now acknowledged that the The sequential transport of cargos between intracellular compart- retromer/WASH complex forms heterogeneous “subcomplexes” ments occurs through multi-step processes that are actively regu- with various interacting proteins to specifically and temporally regu- lated [4]. One important class of regulators controlling vesicular late a large array of cargos [17,20,31–33]. Given the role of RAB

1 Faculté de Médecine et des Sciences de la Santé, Département d’Anatomie et de Biologie Cellulaire, Université de Sherbrooke, Sherbrooke, QC, Canada 2 Center for Biological Systems Analysis (ZBSA), Faculty of Biology, Albert Ludwigs Universitaet Freiburg, Freiburg, Germany *Corresponding author. Tel: +1 819 821 8000; Fax: +1 819 820 6831; E-mail: [email protected]

ª 2019 The Authors EMBO reports 20: 47192 | 2019 1 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

GTPases in membrane traffic, it is likely that other RABs are RAB21. Various RAB21 variants were chosen in order to maximize involved in various aspects of retromer/WASH functions. the recovery of GEFs, GAPs, and effectors. GFP:RAB21 variants RAB interactors have mostly been defined through in vitro pull- were expressed and properly localized in both HeLa and HCT116 down and yeast two-hybrid approaches [34–37]. Pull-down experi- cells, except for T33N that showed weaker early endosomal localiza- ments are a powerful tool to assess direct binding between a RAB tion with a concomitant Golgi relocalization (Fig EV1A–D), consis- and a specific protein, and have been successfully used to identify tent with earlier findings [44]. interactors for multiple RABs [35]. However, they preclude identifi- Duplicate SILAC experiments for each GFP:RAB21 variants iden- cation of context-specific RAB GEFs, GAPs, or effectors, since the tified a vast spectrum of potential direct and indirect RAB21 interac- temporal aspects of effector recruitment are lost. On the other tors (Figs EV1E and EV2A–C, Dataset EV1). Surprisingly, functional hand, yeast two-hybrid approach allows for the identification of annotations in Reactome of individual cell lines did not show an RAB GEFs, GAPs, and effectors. However, yeast two-hybrid enrichment toward membrane trafficking (Fig EV2D and E) and approach only monitors binary interactions, and as a result, were reminiscent of those observed from a recently published GST: complexes interacting with RABs through multiple proteins cannot RAB21-interactome pull-down study (Fig EV2F) [45]. Since the be identified. Recently, a pull-down-based RAB-interactome screen interactome was generated in two cell lines, we hypothesized that was performed in Drosophila enabling identification of a large proteins concomitantly enriched from both cell lines would likely number of RAB effectors [37]. Unfortunately, harder to purify RABs represent RAB21 interactors. Hence, the HeLa and HCT116 RAB21 showed very limited number of interactors [37]. Ongoing networks generated with all of the various RAB21 variants were proteome-wide studies aimed at defining the human proteome [38] merged. This resulted in 29 proteins present altogether (Fig EV1E), have also tested numerous RAB GTPases. Unfortunately, these which interacted to different degrees with the WT, Q78L, or T33N studies used C-terminal tags for affinity purifications, which are not RAB21 variants (Fig EV2C and Dataset EV1). When organized into a appropriate for RABs, due to RAB C-terminal prenylation. The network (Fig EV1F), multiple membrane trafficking regulators or suboptimal tagging of the RABs in these studies yielded a low potential cargos were observed. Reactome pathway analysis number of interactors for most of the twenty-five RABs tested. This revealed an enrichment in membrane and vesicular transport was particularly evident given the strong prevalence for enzymes processes (Fig EV1G), strengthening these core proteins as prospec- linked to RAB prenylation in the interactome (i.e., CHM, CHML, tive RAB21 interactors. While this quantitative AP-MS interactome RABGGTA/B) [38]. Hence, in order to understand how RABs yielded a highly relevant network of RAB21 binding proteins, it was exhibit different cellular functions, it is imperative to accurately nonetheless variable between repeats and cell lines and required the and extensively define their associated proteome in the appropriate use of GDP-/GTP-locked constructs. Therefore, other approaches setting. In an effort to develop new approaches to map RAB were sought to identify RAB regulators and interactors. GTPase interactors, we have combined quantitative mass spectrom- etry and APEX2 proximity labeling techniques. Herein, we describe APEX2:RABs are properly localized in HeLa cells APEX2-mediated proximity labeling as a new highly efficient method to rapidly map RAB regulators/effectors. This approach One caveat of AP-MS and pull-down approaches is that cell lysis notably allowed defining a novel RAB21 interaction with the influences protein–protein interactions. To circumvent this issue, WASH/retromer complexes and a RAB21 role in endosomal sorting APEX2:RAB fusions were used to perform proximity labeling of a subset of clathrin-independent cargos. (Fig 1A) [46]. APEX2 is an engineered peroxidase that biotinylates proteins, mostly on tyrosine [47], in a 10–20 nm radius [48]. It uses biotin–phenol as its substrate, and the reaction is catalyzed by a 1-

Results min H2O2 treatment, enabling the rapid covalent labeling of neigh- boring proteins. Quantitative mass spectrometry defines potential RAB A concern with proximity labeling approaches is that they often interactors lead to the identification of large arrays of proteins [49]. We rationalized that by using three early endosomal RABs as baits, we Early endosomal RAB21 has well-described roles in mediating inte- could (i) validate the technique, (ii) identify general and specific grin internalization to control cell migration, anoı¨kis resistance, and early endosomal neighbors for the three RABs, and (iii) refine the cell division [39,40]. RAB21 also regulates aspects of VAMP7 and RAB21 proteome established by AP-MS. Moreover, given the known VAMP8 trafficking to control neurite growth and autophagy, respec- segregation of RAB5 and RAB4 at endosomes [50,51], we hypothe- tively [41,42]. Unfortunately, a low number of specific interactors sized that APEX2 labeling would also identify RAB-specific microdo- were identified for RAB21 in a recent pull-down study [37]. Given mains (Fig 1B). APEX2, APEX2:RAB21, APEX2:RAB5a, and APEX2: the importance of RAB21-associated functions, and the difficulty in RAB4a HeLa Flp-In/T-REx cells were thus generated. APEX2-fused identifying RAB21 interactors through conventional approaches, RABs were able to biotinylate endogenous proteins. Strong biotiny-

RAB21 represents a good RAB on which to establish novel method- lation was observed in the presence of H2O2 and biotin–phenol, ologies aimed at identifying RAB-associated proteins. Hence, a whereas weak biotinylation was noted when H2O2 was omitted quantitative SILAC-based affinity purification (AP-MS) approach (Fig 1C). Immunofluorescence analyses revealed that biotinylated was devised to map RAB21 interactors. proteins in APEX2:RAB21, RAB5, and RAB4 cells localized on EEA1- The Flp-In/T-REx system [43] was used to generate stable HeLa positive puncta and also in a cytosolic/reticular pattern, while and HCT116 cell lines expressing N-terminally GFP-tagged wild-type APEX2-only cells showed diffuse cytosolic and nuclear staining with (WT), GTP-locked (Q78L), or GDP-locked (T33N) forms of human no EEA1 colocalization (Fig 1D). Together, these results indicate

2 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

B OH 1 min H2O2 B OH treatment B O- B O-

A APEX2 B B APEX2 B B RAB B B RAB APEX2 OH RAB21 EE EE B OH B Early EE Early Early endosome endosome EE endosome B EE B OH B OH RAB4 RAB5 APEX2 B Biotin-phenol APEX2 incubation Biotinylation

C D H O APEX2 2 2 Biotin EEA1 Biotin Zoom H2O2 -+ EEA1 Biotin-phenol + + DAPI

Biotinylated proteins: Biotin EEA1 Biotin EEA1 DAPI

HeLa-APEX2 + APEX2

APEX2: RAB21 Biotin EEA1 Biotin EEA1 H2O2 -+ DAPI Biotin-phenol + +

Biotinylated Biotin EEA1 Biotin proteins: EEA1 DAPI +

APEX2:RAB21 HeLa-APEX2:RAB21

APEX2: Biotin EEA1 Biotin RAB5a EEA1 H2O2 -+ DAPI Biotin-phenol + +

Biotinylated Biotin EEA1 Biotin proteins: EEA1 DAPI +

APEX2:RAB5a HeLa-APEX2:RAB5a

APEX2: Biotin EEA1 Biotin RAB4a EEA1 H2O2 -+ DAPI Biotin-phenol + +

Biotinylated Biotin EEA1 Biotin proteins: EEA1 DAPI +

APEX2:RAB4a HeLa-APEX2:RAB4a

Figure 1.

ª 2019 The Authors EMBO reports 20: 47192 | 2019 3 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

◀ Figure 1.APEX2:RAB expression lead to endosomal biotinylation. A Diagram representing APEX2:RAB-mediated endosomal biotinylation of endogenous proteins. B Illustrative representation of APEX2:RAB endosomal microdomains. C APEX2 only or APEX2:RAB were found to biotinylate endogenous proteins. Streptavidin Western blotting of total biotinylated proteins in APEX2:RAB or APEX2 only Flp-In/T-REx HeLa cells. D APEX2:RAB biotinylated proteins (streptavidin) are partially colocalized with EEA1 in HeLa cells, n = 2 independent experiments. Scale bars: 10 lmor5 lmin enlarged views.

that APEX2 is active when fused with RABs and that APEX2:RAB FAM160A1, and Ccdc128, which represent previously identified can biotinylate proteins at EEA1-positive endosomes. RAB5 effectors [37]. Surprisingly, Rabenosyn-5, which binds both RAB5 and 4, was mostly enriched with RAB5 compared to RAB4. APEX2:RAB proximity labeling is enriched for trafficking RABIP4, RAB11FIP11, and VPS45, which are functionally mostly regulators/effectors linked to RAB4, were also more prominently found with RAB5 or RAB21 compared to RAB4, while only GRASP-1 was notably Three label-free independent proximity labeling experiments for observed with RAB4. Nonetheless, all of these known RAB4 interac- each RABs and the APEX2 control were performed. In order to tors, although less abundant with RAB4, were recovered at an < 1% statistically identify interactors/neighbors and to remove cytosolic/ FDR in APEX2:RAB4 experiments. reticular contaminants, the SAINT software was used for filtering The APEX2 approach also identified known RAB21 and RAB5 and the ProHits-viz suite for data representation [52,53]. Using a GEFs (Fig 2C). VARP, a RAB21 GEF, was highly abundant with very strict SAINT filtering score of ≥ 0.95, which corresponds to a RAB21, while the more general RABGEF1 was observed at a low false discovery rate of ≤ 1%, 1,173 proteins were identified in FDR (< 1%) with RAB21 and 5. RAB GAPs were also recovered by APEX2:RAB21, 819 in APEX2:RAB5, and 469 in APEX2:RAB4 when APEX2:RABs (Fig 2C). Of note, TBC1D2, previously described with normalized to APEX2 only (Fig 2A and Dataset EV2). Importantly, RAB5 GAP activity in C. elegans [56], was abundant with RAB5, biotinylation was equivalent between APEX2:RABs (Fig EV3A). while GAPCENA, a known RAB4 GAP, was present with RAB4 [57]. Therefore, the different number of interactors/neighbors between No RAB21 GAPs have been described in the literature, although the various RABs most probably reflects different degrees of proxim- TBC1D17 was shown to have weak catalytic activity toward RAB21 ity with endosomal proteins, rather than variable protein expression in vitro [57]. TBC1D4 and/or TBC1D15, a close relative of TBC1D17, levels between baits. RAB5 and 21 showed 57% overlap between could potentially act on RAB21 given their respective high abun- their neighboring proteins (666 proteins), while RAB21/RAB4 and dance with RAB21. RAB5/RAB4 had 35% (412 proteins) and 51% overlap (415 To confirm that APEX2:RAB-identified proteins were recovered proteins), respectively (Fig 2A). When all RAB21, RAB5, and RAB4 due to their close proximity to RABs, and not merely the result of neighbors were analyzed for gene ontology (GO) term enrichment, their general endosomal localization, the RAB21 dataset was molecular functions related to trafficking events were highly signifi- compared to a previously published APEX2:2xFYVE dataset [58]. cant and overlapped extensively (Fig 2B). Of significance, the The data were normalized using ProHits-viz and compared. After APEX2:RAB approach highlighted cellular functions related to traf- analysis, most proteins were only found with RAB21 (Fig EV3B), ficking, something that both GST pull-down and AP-MS approaches while those shared between the 2xFYVE probe and RAB21 were, for failed to achieve without heavy filtering and the use of numerous the most part, known PtdIns(3)P binding or associated proteins RAB variants and cell lines. (Fig EV3C). Thus, proteins identified through APEX2:RAB were not simply observed due to their endosomal localization. To further Wild-type APEX2:RABs identify known GEFs, GAPs, and effectors ensure the specificity of the APEX2:RAB-identified proteins, early endosomal RABs were compared to APEX2:RAB7a, a late endoso- Given the large number of proteins identified in the APEX2:RAB21, mal RAB. APEX2:RAB7 localized appropriately and led to biotinyla- 5, and 4 datasets, the type of neighbors detected through this tion at late endosomes/lysosomes as detected with LAMP1 methodology was explored. A comparative analysis of the various (Fig EV3E). APEX2:RAB7-identified proteins had a limited overlap RABs (Fig 2C) on ProHits-viz was performed. Strikingly, GEFs, with early endosomal RABs as expected (Figs 2A–C and EV3D). GAPs, and effectors/interactors were identified from wild-type RABs Importantly, the well-described RAB7 GAP TBC1D5 [14] and with mostly specific enrichments toward their predicted target. the known RAB7 interactor VPS35 [59] were highly enriched with Using published interactors (without exhaustively listing them all, RAB7 (Fig 2C). Altogether, these experiments strengthen APEX2: Fig 2C and Dataset EV2), preferential associations between RAB21 RAB proximity labeling as an efficient method to identify RAB and MTMR2 were observed, as shown in Drosophila [54] and with interactors/neighbors. VAMP7 [41]. An association between RAB21 and integrin(s) was Another important aspect of the APEX2:RAB proximity labeling also detected, although integrins were rather variable in their prox- approach was that it successfully enriched whole protein complexes imity with early endosomal RABs, as could be expected from previ- with single RABs (Fig 2C). This approach, using a single purification ous reports [55]. APPL1/2 and EEA1, which are known RAB21 and step, effectively identified all subunits of the EARP complex with RAB5 effectors, were also highly recovered with both RAB21 and RAB4. This complex was recently identified as an early endosomal RAB5, while weak in RAB4. APEX2:RAB5-mediated proximity liga- sorting complex interacting with RAB4 [60,61]. EARP is highly simi- tion also identified other known RAB5 effectors, including VPS34 lar to the Golgi GARP complex, which does not interact with RAB4 and KIF13A. It also enriched HOOK1/3, UHRF1BP1L, CC2D1A, [37]. Of note, all EARP-specific subunits were identified with strong

4 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

ABRAB5 RAB4 membrane organization intracellular protein transport

113 17 ER to Golgi vesicle-mediated transport

protein transport

0 40 36 vesicle-mediated transport 32 COPII vesicle coating 00 355 468 cell division retrograde transport, endosome to Golgi 0 20 284 mitosis RAB7 17 RAB21 post-Golgi vesicle-mediated transport 2 -Log10(Pvalue) 0 2 4 6 810121416 RAB21 RAB5 RAB4 RAB7

C Effectors GEFs Sorting complexes RAB21 RAB21 RAB21 RAB5 RAB4 RAB7 RAB5 RAB4 RAB7 RAB5 RAB4 RAB7

MTMR2 VARP CCDC132 RAB21 RAB21 VAMP7 RABGEF1 VPS52 EARP ITGB1 GAPVD1 RAB5 complex VPS53 RAB4 ITGB4 Ang2 ITGA3 RAB21 TSSC1 and GAPs ITGAV RAB5 Strumpellin RAB21

APPL2 RAB5 RAB4 RAB7 KIAA1033 APPL1 WASH FAM21A EEA1 Complex WASH TBC1D4 Potential Rabankyrin 5 TBC1D15 RAB21 CAPZA2 VPS15 TBC1D2 RAB5 CCDC53 VPS34 GAPCENA RAB4 VPS35 KIF13A TBC1D9B VPS26A HOOK1 RAB5 TBC1D5 VPS26B HOOK3 retromer TBC1D23 VPS29 UHRF1BP1L SNX3 CC2D1A SNX1 FAM160A1 Retromer SNX2 Ccdc128 07AvgSpec 5Complex SNX5 Rabaptin-5 SNX6 Rabenosyn-5 RAB5 and SNX27 RABIP4 Relative abundance RAB4 DNAJC13 RAB11FIP1 RAB7A VPS45 RAB4 ≤ 0.01 ≤ 0.05 > 0.05 FKBP15 GRASP-1 BFDR EHD1

Figure 2. Comparison of APEX2:RAB21-, 5-, 4-, and 7-generated proteomes. A Venn diagram highlighting the number of interactors with each RAB and their overlap. B Reactome analysis of all APEX2:RAB21, APEX2:RAB5, APEX2:RAB4, and APEX2:RAB7 interactors/neighbors. C ProHits-viz generated dot plots of RAB effectors, GEFs, GAPs, and sorting complexes. Green, blue, and beige shadings along with RAB names refer to previously identified interactions between RABs and the depicted proteins.

ª 2019 The Authors EMBO reports 20: 47192 | 2019 5 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

enrichment with RAB4, while the GARP complex-specific subunit anti-biotin immunoprecipitation [67] on APEX2:RAB-expressing was not recovered. Also, the WASH [62] complex was strongly cells, and the presence of WASH- and retromer-associated proteins enriched with RAB21, while a few subunits were observed with was tested. Using this approach, Strumpellin was more prevalent RAB5 and RAB7 (Fig 2C). Interestingly, VPS retromer subunits [18] with APEX2:RAB21, while similar amounts of VPS35 between RAB5 were strongly enriched with RAB7, compared to early endosomal and RAB21 (Fig 3A) were detected. This trend is in accordance with RABs (Fig 2C), as expected from previous work [13,14]. Although the mass spectrometry approach (Fig 2C). early endosomal RABs were statistically enriched with the VPS Specific interactions between RAB21 and WASH/retromer retromer subunit, they showed stronger proximities to various SNXs complexes subunits were further tested by CoIP. Endogenous and other retromer subunits (i.e., SNX2, SNX6). VPS26, FAM21, Strumpellin, and VPS35 immunoprecipitated with RAB GTPases are known to be functionally linked to phospho- GFP:RAB21 (Fig 3B and C). Although weak, these interactions were inositides [5]. Thus, their proximity to phosphoinositide regulatory highly reproducible and associated with different degrees with the enzymes was investigated. RAB21 and RAB5 were found to interact various RAB21 variants. This is similar to RAB21 association with or be in close proximity to a wide range of phosphoinositide phos- integrins [68]. Of particular note, VPS35 was also observed in the phatases and kinases (Fig EV4A). In accordance with previous stud- AP-MS SILAC experiments (Dataset EV1). Lastly, pull-down experi- ies [63], APEX2 proximity labeling identified INPP4a and OCRL as ments on HeLa cell lysates with bacterially purified GST:RAB21 [54] strong RAB5 interactors/neighbors. Interestingly, multiple Myotubu- were performed to assess a more direct interaction between the larin family members were specifically identified with RAB21. More- WASH/retromer complexes and RAB21. FAM21, Strumpellin and over, PIKFYVE was also exclusively found in APEX2:RAB21 cells CAPZa, three WASH complex components, and VPS35, a core (Fig EV4A). These associations imply that RAB21 could potentially retromer subunit (Fig 3D), were detected. The efficiency of the pull- impact three phosphoinositide pools, namely PtdIns(3)P, PtdIns(5) down was confirmed by the presence of APPL1 (Fig 3D). While

P, and PtdIns(3,5)P2. Altogether, APEX2:RAB proximity labeling these pull-down results do not allow concluding that RAB21 directly allowed identifying a wide array of known regulators/interactors of interacts with the WASH or retromer complexes, they nevertheless RAB GTPases. strongly suggest that RAB21 binds to these complexes either directly or indirectly. Finally, as expected from the interaction data, partial APEX2:RABs identify novel RAB interacting proteins colocalization between RAB21 and the WASH and retromer complexes was observed (Fig 3E and F), further indicating that Most of the aforementioned proteins represent known interactions RAB21 interacts with the WASH and retromer complexes. and complexes. Consequently, the extent of the interactors/neigh- bors identified by the APEX2 technique validated the APEX2 RAB21 is required for complete endosomal WASH/retromer approach. However, in addition, we also assessed whether APEX2: complex recruitment RABs proximity labeling could identify novel regulatory interac- tions. In order to identify RAB21 interactors/neighbors, APEX2: To functionally assess whether RAB21 affects WASH/retromer func- RAB21 enriched baits were filtered (Fig EV4B). This led to the iden- tions, polyclonal RAB21 knockout HeLa cell populations were tification of (i) PLEKHM2, which has known roles in lysosomal posi- generated using CRISPR/Cas9 [69]. To ensure the specificity of the tioning and autophagy [64], (ii) SLC7A11, which interacts with assayed phenotypes, two cell populations using independent guide SLC3A2 to control amino acid transport [65], and (iii) USP7, a RNAs targeting distinct genomic regions were generated. Thus, simi- recently identified WASH complex modulator [66]. Other identified lar phenotypes in both cell populations would strongly argue for a RAB21 interactors/neighbors showed potential roles in TGF-b,in specific effect. RAB21 knockout was confirmed through sequencing adherent junctions, and in cohesin functions. To assess whether of the targeted genomic regions, with various indels or insertions proteins detected by mass spectrometry could be validated by other observed for each gRNA in the polyclonal populations (Fig EV4E). approaches, APEX2 and APEX2:RAB21 proximity labeling in GFP: Some non-edited cells were observed in the gRNA-2 cell population SARA transfected cells was performed. After cell lysis and GFP: compared to the gRNA-3 population (Fig EV4E), which correlated SARA immunoprecipitation, biotinylated SARA was only observed with RAB21 protein levels detected by Western blotting (Fig 4A). in APEX2:RAB21 cells (Fig EV4C). Moreover, co-immunoprecipita- Importantly, as expected from previous studies [42], the number of tion (CoIP) experiments performed between FLAG:RAB21WT- and LC3 puncta was increased in the two RAB21 knockout cell popula- HA:PLEKHM2- or HA:SLC7A11-expressing cells allowed highlight- tions (Fig EV4F). Morphological analysis of various membrane ing reproducible interactions between RAB21 and PLEKHM2 or compartments did not identify major defects in compartment size SLC7A11 (Fig EV4D). These examples further strengthen the APEX2 and localization (Fig 4B), aside from transferrin uptake being slower proximity labeling technique and corroborate these newly estab- in RAB21 KO cells (Fig EV4G and H) in accordance with previous lished neighbors as novel RAB21 interactors. work [44]. RAB21 knockout led to a slight reduction in VPS35 and VPS26 RAB21 interacts and colocalizes with the WASH and protein levels, while it did not affect VPS29. On the opposite, WASH, retromer complexes Strumpellin, and CAPZa proteins levels showed a slightly increasing trend (Fig 5A and B), indicating that RAB21 may regulate the local- In order to further build from the APEX2 proximity labeling ization or stability of these complexes. Significantly, loss of RAB21 approach and define novel molecular functions for RAB21, its prox- reduced VPS35 recruitment to endosomes detected both by a imity to the WASH and retromer complexes was studied in further decreased number of VPS35 puncta per cells and by a reduced colo- detail. Mass spectrometry data were confirmed by performing an calization between VPS35 and EEA1 (Fig 5C–E). Decreased

6 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

AB C APEX2:RABs GFP:RAB21 GFP:RAB21

APEX2 4 521 FT

WT Q78L T33N FT WT Q78LT33N part PFG part Strumpellin VPS26 FAM21 VPS35

GFP:RAB21PFG part I.P biotin I.P Strumpellin Strumpellin Lysates VPS26 VPS35 VPS35 GFP Lysates

Strep FAM21 setasyL Strumpellin

DEGST Pull-Down VPS35 Input

2% GST RAB21 GFP:RAB21 GFP:RAB21 GFP:RAB21 EEA1 Strumpellin EEA1 EEA1 DAPI DAPI FAM21 CAPZa VPS35 EEA1 APPL1 GFP:RAB21 GFP:RAB21 GFP:RAB21 FAM21 GST:RAB21 FAM21 FAM21 DAPI DAPI Ponceau

GST FAM21

F 0.4 GFP:RAB21 GFP:RAB21 GFP:RAB21 WASH WASH WASH DAPI DAPI 0.3 WASH 0.2 GFP:RAB21 GFP:RAB21 GFP:RAB21 VPS26 VPS26 VPS26 0.1 DAPI DAPI per cell RAB21 vs markers Average pearson correlation 0 n≥ VPS26 23 cells EEA1 FAM21 WASHVPS26

Figure 3. RAB21 interacts and colocalizes with the WASH and retromer complexes. A Strumpellin and VPS35 are in close proximity to RAB5 and RAB21. Anti-biotin immunoprecipitation and immunoblot of endogenous Strumpellin and VPS35. Lysates correspond to 1% of input, n = 3 independent experiments. B, C RAB21 can be seen interacting with various WASH and retromer complexes subunits. GFP-Trap IP of WT, Q78L, and T33N RAB21 variants in HeLa cells followed by GFP immunoblot and endogenous (B) VPS26 immunoblot and (C) FAM21, Strumpellin, and VPS35 immunoblots. Lysates correspond to 5% of input, n ≥ 3 independent experiments. D RAB21 actively pulls down WASH and retromer complex subunits. Bacterially purified and GTP-loaded GST:RAB21 pull-down of HeLa cell lysates followed by Strumpellin, FAM21, CAPZa, and VPS35 immunoblots. Ponceau staining reveals the purity of the GST and GST:RAB21 used for the pull-downs, n = 3 independent experiments. E RAB21 colocalizes with the WASH and retromer complexes. Transiently expressing GFP:RAB21 HeLa cells were fixed and stained for endogenous EEA1, FAM21, WASH, and VPS26. Boxed region is magnified, and single channels are depicted. Scale bars: 10 lmor5 lm in enlarged view. F Pearson’s correlation per cell between RAB21 and the various markers. Error bars are SEM, n = 2 independent experiments.

ª 2019 The Authors EMBO reports 20: 47192 | 2019 7 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

Early Endocytic Recycling Late Lysosomes ABEndosomes compartments Endosomes APPL1 TfR RAB7 LAMP1 KO DAPI DAPI DAPI DAPI

Parental gRNA-2 gRNA-3 RAB21 Parental RAB5 RAB7 APPL1 TfR RAB7 LAMP1 DAPI DAPI DAPI DAPI GAPDH gRNA-2

APPL1 TfR RAB7 LAMP1 DAPI DAPI DAPI DAPI gRNA-3

Figure 4. Characterization of RAB21 knockout HeLa cells. A Low RAB21 expression levels in two independent RAB21 KO cell populations. Immunoblot analysis of endogenous RAB21, RAB5, RAB7, and GAPDH. B Endo-lysosomal compartments are unaffected by the loss of RAB21. Immunofluorescence of APPL1, TfR, RAB7, and LAMP1 in parental HeLa cells and in the two RAB21 knockout cell populations. Scale bars: 10 lm, n = 3 independent experiments.

colocalization between SNX1 and EEA1 was also observed, although sorting, respectively [17,71], the localization of these two proteins the number of endosomal SNX1 puncta was not affected by RAB21 was therefore assessed. Surprisingly, in RAB21 knockout cells, no deletion (Fig EV5A–C), suggesting that the cargo sorting complex is defects were observed for either cargos (Fig 6A–C), thereby indi- more affected by the loss of RAB21 than the sorting nexin complex. cating that RAB21 is not required for general retromer sorting WASH and FAM21 also showed, albeit to a weaker extent than events, and consistent with the observed partial endosomal loss of VPS35, decreased endosomal colocalization with EEA1 (Figs 5F–H WASH/retromer complexes (Fig 5C–H). Given the recently and EV5D). Altogether, these results demonstrate that RAB21 is reported heterogeneity of retromer complexes [20], it is possible required for the full recruitment of the WASH and retromer that RAB21 modulates the sorting of a defined set of cargos. complexes at EEA1-positive endosomes. To assess whether RAB21 Indeed, VARP was recently identified as a retromer interacting was required for a specific WASH-mediated process, endosomal F- protein and shown to be involved in regulating a subset of VPS35- actin was monitored since it is believed to be mostly controlled by dependent cargos. Since RAB21 KO did not impact CI-MPR or the WASH complex [27,28,70]. Importantly, endosomal F-actin was GLUT1 trafficking, the VARP-dependent cargo MCT1 was assessed. decreased in RAB21 KO cells compared to controls (Fig 5I and J) RAB21 KO notably affected MCT1 protein levels (Fig 6D and E). demonstrating the importance of RAB21 for proper WASH function. This defect is reminiscent of what was observed in VARP-depleted This was not a global loss of F-actin, since cortical actin was present cells [20], thus indicating that RAB21 modulates a subset of equally in parental and RAB21 knockout cells (Fig 5I). retromer cargos. Interestingly, MCT1 (SLC16A1) was present in both GFP-Trap RAB21 regulates trafficking of a subset of retromer cargos and APEX2:RAB21 datasets. Since SLC3A2, Basigin (CD147), and SLC16A3 were also observed in both datasets (Datasets EV1 and The interaction and impact of RAB21 on WASH/retromer localiza- EV2), this could indicate that their trafficking may be regulated by tion and actin polymerization suggest that RAB21 may regulate RAB21, similar to MCT1. The interaction between RAB21 and the trafficking of a larger proportion of cargos than initially SLC3A2 (Fig 6F) was thus first validated. Confirming the proteomic believed [39,42]. WASH/retromer complexes regulate endosome- data, FLAG:RAB21 efficiently immunoprecipitated HA:SLC3A2. to-Golgi retrograde pathways and cargo recycling between endo- Given the available reagents, only SLC3A2 and Basigin trafficking somes and plasma membrane [17,29]. Thus, given that both were subsequently investigated. Importantly, these two cargos CI-MPR and GLUT1 proteins represent two well-defined retromer represent “direct CIE cargos” [72], which are characterized by their cargos undergoing retrograde or endosome-to-plasma membrane rapid sorting into tubular endosomes in HeLa cells [73]. As a result

8 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

1.5 3 WASH FAM21 Strump. CAPZa VPS35 VPS26 VPS29 A B ParentalgRNA-2gRNA-3 ParentalgRNA-2gRNA-3 1.0 Strumpellin 2 VPS35 * WASH

VPS26 0.5 FAM21 / GAPDH 1 VPS29 CAPZa GAPDH complex Ratio WASH 0 GAPDH 0 Ratio retromer VPS / GAPDH Ratio retromer RAB21 n=4 RAB21 n=4 gRNA-2 ParentalgRNA-3ParentalgRNA-2gRNA-3ParentalgRNA-2gRNA-3 ParentalgRNA-2gRNA-3ParentalgRNA-2gRNA-3ParentalgRNA-2gRNA-3ParentalgRNA-2gRNA-3 C VPS35 / EEA1 / DAPI DE<0.0001 0.5 VPS35 VPS35 VPS35 EEA1 <0.0001 60 EEA1 EEA1 DAPI DAPI 0.4 40 0.3 Parental

0.2 20

0.1 Average number of VPS35 objects per cell per cell VPS35 vs EEA1

Average pearson correlation 0 0 ≥ ≥

gRNA-2 n n 22 cells 22 cells

gRNA-2 gRNA-3 gRNA-2 gRNA-3 Parental Parental

F WASH / EEA1 / DAPI GH WASH WASH WASH EEA1 EEA1 EEA1 0.5 0.6 DAPI DAPI 0.4 0.4 Parental 0.3

0.2 0.2 0.1 per cell WASH vs EEA1 per cell FAM21 vs EEA1 Average pearson correlation Average pearson correlation

gRNA-2 0 0 n ≥ n ≥ 40 cells 43 cells

ParentalgRNA-2 gRNA-3 ParentalgRNA-2 gRNA-3 I F-Actin / EEA1 / DAPI J F-actin F-actin F-actin EEA1 0.10 EEA1 EEA1 <0.0001 DAPI DAPI 0.08

0.06 Parental

0.04

0.02 per cell F-actin vs EEA1

Average pearson correlation 0

gRNA-2 n ≥ 67 cells

ParentalgRNA-2 gRNA-3 Figure 5.

ª 2019 The Authors EMBO reports 20: 47192 | 2019 9 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

◀ Figure 5. RAB21 modulates WASH and retromer endosomal recruitment and WASH activity. A The stability of the retromer cargo sorting complex is affected in RAB21 KO cells. Immunoblot analysis of VPS35, VPS26, VPS29, GAPDH, and RAB21 in parental and the two RAB21 KO cell populations. Ratio of VPS35, VPS26, and VPS29 integrated densities to GAPDH of four independent experiments. *P < 0.05. B WASH complex proteins are not affected by RAB21 deletion. Immunoblot of Strumpellin, WASH, FAM21, CAPZa, GAPDH, and RAB21. Ratio of WASH, FAM21, Strumpellin, and CAPZa integrated densities to GAPDH of four independent experiments. C–E RAB21 knockout decreases VPS35 localization at endosomes. (C) Immunofluorescence of endogenous VPS35 and EEA1 in RAB21-depleted cells. Boxed region is magnified, and single-channel images are depicted. (D) Pearson’s correlation per cell between VPS35 and EEA1, n = 3 independent experiments. (E) Average number of VPS35 puncta per cell, n = 3 independent experiments. F–H RAB21 loss impairs endosomal recruitment of the WASH complex. (F) Immunofluorescence of endogenous WASH and EEA1 in RAB21-depleted cells. Boxed region is magnified, and single-channel images are depicted. (G) Pearson’s correlation per cell between VPS35 and EEA1 and (H) FAM21 and EEA1, n = 2 independent experiments. I, J RAB21 knockout reduces endosomal F-actin levels. (I) Immunofluorescence of endogenous F-actin using Alexa 488-conjugated phalloidin and EEA1 in RAB21- depleted cells. Boxed region is magnified, and single-channel images are depicted. (J) Pearson’s correlation per cell between internal F-actin and EEA1,n= 3 independent experiments. Data information: In (C, F, I), scale bars represent 10 lmor2.5 lm in enlarged views. Statistical tests used are as follows: (A and B) one-sample t-tests, (D) unpaired t-tests, (E and J) Mann–Whitney tests. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats.

of this feature, their tubular endosomal localization was monitored a common trafficking pathway essential for both SLC3A2 and through a well-defined antibody uptake assay [74]. As expected, Basigin sorting and requiring RAB21, WASH, and the retromer. SLC3A2 and Basigin both localized to endosomal tubules in control RAB21 interaction with the WASH/retromer complexes was inde- cells (Fig 6G and H). Conversely, in RAB21 KO cells, both failed to pendent of RAB21 GTP status (Fig 3C), thus suggesting that these reach tubular endosomes and were observed in a vesicular pattern complexes are unlikely to act as RAB21 effectors. Hence, it is possi- (Fig 6G and H). Importantly, SLC3A2 presence on tubules was ble that the retromer may additionally regulate RAB21 activity, partially rescued by transient overexpression of FLAG:RAB21 WT given its association, through VPS29, with VARP, a RAB21 GEF (Fig EV5E). CD44, another well-defined “direct CIE cargo”, which [20,32,76]. This would provide a potential feedforward, or amplify- was not observed in either the AP-MS or APEX2 datasets, was also ing loop, to ensure concomitant RAB21 activation and retromer/ followed in order to monitor whether RAB21 only affected interact- WASH recruitment at specific endosomal cargos. To assess this, we ing cargos or rather acted more generally in tubular endosome sort- monitored RAB21 endosomal localization in VPS29 and VARP KO ing events. CD44, similar to SLC3A2 and Basigin, failed to reach cells. Interestingly, colocalization between RAB21 and EEA1 or endosomal tubules in RAB21 KO cells (Fig 6G and H). These find- VPS26 (Fig 7E–G) decreased in KO cells. Overall, these colocaliza- ings thus suggest a more general role of RAB21 in mediating fast tion studies suggest that the retromer also regulates RAB21 func- endosomal sorting of “direct CIE cargos”. tions, most likely through its interaction with VARP. Finally, to further corroborate that the loss of RAB21 affects the sorting of these cargos, steady-state SLC3A2 protein levels were followed, with the prediction that the latter would decrease due to Discussion lysosomal degradation, as observed for many retromer-dependent cargos [17]. In accordance with a sorting defect, an increased colo- Defining a global picture of RAB GTPase interactors/neighbors has calization between SLC3A2 and late endosomes (RAB7) and lyso- been hampered by inherent limitations related to the techniques somes (LAMP1; Fig EV5F and G) was detected. Moreover, SLC3A2 used. Through the use and comparison of AP-MS and APEX2 prox- total protein levels were decreased in RAB21 KO cells, as assessed imity labeling, the present data show that APEX2:RAB proximity by both FACS and immunoblotting (Fig EV5H and I). Altogether, labeling was efficient at identifying RAB interacting/neighboring these findings illustrate a new role for RAB21 in controlling endoso- proteins ranging from GEFs, GAPs, effectors, and protein mal sorting of direct clathrin-independent cargos. complexes. While important novel interactions were uncovered through the AP-MS technique, the APEX2 approach outperformed Both WASH and retromer complexes are required for SLC3A2 and the former in terms of specificity and coverage. In addition, a new Basigin sorting and for full RAB21 activity link between RAB21 and the WASH and retromer complexes was established from analysis of the proteomic datasets. As such, RAB21 To confirm that SLC3A2 and Basigin sorting to endocytic tubules was found to be required for complete recruitment of the WASH require the WASH and retromer complexes, knockout HeLa cell and retromer complexes to endosomes and for WASH-mediated populations for FAM21 (WASH), VPS29 (retromer), and VARP actin polymerization. Moreover, RAB21 deletion compromised the (retromer subcomplex; Fig 7A) were generated. Of note, both trafficking of specific clathrin-independent cargos. This impairment complexes were needed for full sorting of SLC3A2 and Basigin into was also observed in retromer or WASH complex-deleted cells. In endosomal tubules (Fig 7B and C). Although the effect was weaker light of the above, it is proposed that RAB21 and the retromer/ compared to RAB21 (Fig 6H), there was a clear drop in the number WASH complexes operate in a feedforward loop to ensure their of cells harboring SLC3A2- and Basigin-labeled tubules in FAM21, respective appropriate endosomal recruitment in order to mediate VPS29, and VARP knockout cells (Fig 7B and C). Importantly, efficient cargo sorting (Fig 8). SLC3A2 total protein levels were also downregulated in WASH and By comparing the neighboring proteomes of three early endoso- retromer knockout cells (Fig 7A and D). This latter finding is also in mal RABs, a strong overlap between the latter was observed, in accordance with previous proteomic studies [17,75] and argues for addition to specific enriched neighbors for each of these RABs, in

10 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors oa e loe al et Olmo Del Tomas ª 2019 represent Figure fclspeetdo ahgahrpeet h oa ubro el nlzdfralrepeats. all for analyzed cells of number total the represents graph each on presented cells of A– ,EMCT E D, aaifrain n() cl asrepresent bars scale (A), In information: Data ,HRAB H G, RAB F RAB C h Authors The oe eini anfe.()Pearson (B) magnified. is region Boxed needn experiments. CD 147 independent for efficient fully not was Alexa eaiefursec unit), fluorescence relative needn experiments. independent 6 RAB21 . 20 21 21 21 1 488 rfikn eursRAB21 requires trafficking sntrqie o IMRo GLUT or CI-MPR for required not is srqie o prpit SLC appropriate for required is SLC with interacts l mor cnuae nimueatbd.Arwed on oedsmltbls oeta o-nenlzdatbde eermvdb nai ah which wash, acid an by removed were antibodies non-internalized that Note tubules. endosomal to point Arrowheads antibody. anti-mouse -conjugated G DE AB srqie o nooa otn fseii ag types. cargo specific of sorting endosomal for required is

5 CD44 Basigin SLC3A2 MCT1 GLUT1 CI-MPR GLUT1 DAPI LAMP1 DAPI SLC3A2 DAPI MCT1 CD44 DAPI Basigin DAPI TGN46 CI-MPR DAPI l nelre iw.Saitcltssue r sflos B unpaired (B) follows: as are used tests Statistical views. enlarged in m aetlgN- gRNA-3 gRNA-2 Parental aetlgN- gRNA-3 gRNA-2 Parental aetlgN- gRNA-3 gRNA-2 Parental RAB 3A2 21 n 21 FLAG:RAB of immunoprecipitation FLAG . = euae I ag sorting cargo CIE regulates 2 D muoloecneo noeosMCT endogenous of Immunofluorescence (D) . n CD44 and needn experiments. independent 3A2 ’ orlto e elbtenC-P n TGN and CI-MPR between cell per correlation s 20 aii,adCD and Basigin, , 1 l xliigpam ebaelbln.()Pretg fclswt uue nprna rRAB21 or parental in tubules with cells of Percentage (H) labeling. membrane plasma explaining , mor rfikn.()Imnfursec fedgnu IMRadTGN and CI-MPR endogenous of Immunofluorescence (A) trafficking. 5 l nelre iw o IMR hl hyrepresent they while CI-MPR, for views enlarged in m 44 rfikn.()Atbd paeasy nprna rkoku RAB21 knockout or parental in assays uptake Antibody (G) trafficking. n muolto oepesdSLC co-expressed of immunoblot and 1 nwl-yeo RAB or wild-type in H

30 cells Average MCT1 integrated n ≥ 46 71 cells 76 cells t intensity per cells (RFU) tssad( n )Mann H) and (C and -tests n ≥ n ≥

Percentage of cells with tubules 1000 2000 3000 4000 Average pearson correlation r()GLUT (C) or

20 40 60 80 per field CIMPR vs TGN46 0.1 0.2 0.3 SLC3A2 0

Pare 0 Parental Parental 0 ntal gRNA-2 <0.0001 gRNA-2 gRNA-2 21 n.s. 1 gRNA-3 dpee el.()Itgae MCT Integrated (E) cells. -depleted 1 LAMP and gRNA-3 gRNA-3 Basigin

10 Parental <0.0001 3A2 l gRNA-2 m C 46 H.Lstscrepn to correspond Lysates :HA. , nelre iw o GLUT for views enlarged in gRNA-3 n F – rGLUT or hte et.Alerrbr r E.Number SEM. are bars error All tests. Whitney = MOreports EMBO

3 Parental 86 cells Lysates CD44

needn experiments. independent Average pearson correlation n ≥ <0.0001

1 gRNA-2 per field GLUT1 vs LAMP1 0.005 0.1 0.2 0.3 0.4 0.5 ihLAMP with SLC3A2:HA gRNA-3 I.P. FLAG Parental 0

el.Clswr tie ihan with stained were Cells cells. FLAG gRNA-2 n.s. 20 1 : 1 FLAG:RAB21 gRNA-3 47192 nRAB21 in nest e el(in cell per intensity HA FLAG HA input, 5% 1 ncotcells, knockout . (D | 2019 n )Saebars Scale G) and ncotcells. knockout MOreports EMBO n = 3 11 n = of 3 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

ABParental ∆FAM21 ∆VARP ∆VPS29 SLC3A2 DAPI

FAM21 VARP VPS29 ∆ ∆ ∆ Parental FAM21 VARP VPS29 Basigin DAPI SLC3A2 GAPDH Basigin SLC3A2

C D 0.0094 E Parental ∆VARP ∆VPS29 0.0058 <0.0001 0.0458 RAB21 0.0021 EEA1 80 1.0 DAPI 0.0002 <0.0001 0.8 60 0.6 40 0.4 20 0.2 Average percentage of Average Ratio SLC3A2 / GAPDH cells with tubules per field cells with tubules 0 0

n ≥ GFP:RAB21 / EEA1 DAPI 217 cells n = 3 VARP VARP VARP FAM21∆ VPS29 FAM21∆ ∆ VPS29 FAM21∆ VPS29 Parental ∆ Parental∆ ∆ Parental∆ ∆ SLC3A2 Basigin G F Parental ∆VARP ∆VPS29 0.046 RAB21 0.4 0.057 0.25 <0.0001 VPS26 DAPI

0.3 0.20

0.15 0.2 VPS26 / DAPI 0.10 0.1 0.05 Average pearson correlation Average pearson correlation per cell GFP:RAB21 vs EEA1

0 per cell GFP:RAB21 vs VPS26 0

n ≥ n ≥ GFP:RAB21 / 23 cells VARP 37 cells VARP ∆ VPS29 ∆ VPS29 Parental ∆ Parental ∆

Figure 7. WASH and retromer complexes are required for SLC3A2 and Basigin sorting, and for full RAB21 activation and endosomal localization. A Validation of WASH and retromer knockouts and SLC3A2 protein levels. Immunoblots of parental or FAM21, VARP, and VPS29 knockout cell populations. B Antibody-uptake assays in parental or knockout FAM21, VARP, and VPS29 cells. Cells were stained with an Alexa 488-conjugated anti-mouse antibody. Arrowheads point to endosomal tubules. C Percentage of cells with tubules in parental or in various knockout cells; n=3 independent experiments. D Ratio of SLC3A2 integrated densities to GAPDH; n = 3 independent experiments. E–G Retromer is required for endosomal localization of RAB21. Immunofluorescence of transiently expressed GFP:RAB21 with endogenous (E) EEA1 or (F) VPS26 in parental or VARP- and VPS29-deleted cells. Boxed region is magnified underneath. (G) Pearson’s correlation per cell between GFP:RAB21 and EEA1 or VPS26; n = 3 independent experiments. Data information: (B) Scale bars represent 20 lm, (E and F) scale bars represent 10 lmor2.5 lm in enlarged views. Statistical tests used are as follows: (C) unpaired t-tests, (D) one-sample t-tests, (G) Mann–Whitney tests. All error bars are SEM. Number of cells presented on each graph represents the total number of cells analyzed for all repeats.

12 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

Clathrin independent WT cells endocytosis RAB21 KO cells

Recycling tubular endosomes

actin

RAB21 Recycling Recycling

VARP endosome endosome

Lysosomal mis-routing

Anterograde Anterograde trafficking trafficking

Direct CIE cargos (SLC3A2, Basigin, CD44, Mct1(?) Glut1 CI-MPR WASH Complex Retromer

Figure 8. Model of RAB21-mediated WASH/retromer cargo sorting. RAB21 is required for endosomal sorting of direct clathrin-independent cargos. RAB21 associates with WASH/retromer subcomplexes at endosomes. RAB21 could either (i) recruit WASH/retromer or (ii) be recruited by WASH/retromer or (iii) be part of a positive feedback loop that would allow WASH/retromer and RAB21 recruitment at endosomes. Endosomal RAB21 would be required for WASH-mediated F-actin polymerization. Although the data do not directly demonstrate a direct link between F-actin generation and cargo sorting, we propose that RAB21-dependent F-actin generation would be required for sorting of a CIE cargo subclass (MCT1, SLC3A2, Basigin, and CD44), while it would not be required for other cargos (CI-MPR or Glut1). In RAB21 knockout cells, decreased WASH/retromer endosomal localization is observed, which results in reduced endosomal F-actin and direct CIE cargo misrouting. This ultimately leads to the lysosomal degradation of these misrouted cargos (demonstrated for SLC3A2). Dotted line between VARP and RAB21 indicates that their coregulation with the retromer is speculative.

accordance with the concept of RAB microdomains [50,51]. These direct comparison of related RABs yielded more valuable informa- observations support the use of APEX2 as an efficient tool to identify tion than an ectopically endosomal-enriched probe. Moreover, when RAB regulators/effectors. One caveat however with APEX2 is that it using a previously published APEX2:2xFYVE dataset [58], we does not discriminate between direct and indirect interactors, nor observed that certain RAB interactors would have been filtered out, with bystander proteins. In general, proximity labeling often leads since they are also associated with PtdIns(3)P binding proteins (i.e., to the identification of a large quantity of proteins [49]. Indeed, EEA1 [77]). Another approach that could define proteins more there are a large number of statistically enriched proteins for all directly associated with RABs would be to use anti-biotin immuno- RABs, thus rendering their follow-up study prioritization difficult. precipitation of trypsin-digested APEX2:RAB lysates in order to Over 1,375 significant proteins were identified herein when combin- directly map biotinylated peptides [67]. Comparing the latter over a ing four RABs. While they most probably not all represent direct general streptavidin enrichment would in principle allow mapping interactors or direct RAB/protein complexes, the present data indi- of more direct interactions, as recently shown [78], and would most cate that they reflect the protein environment associated with each probably significantly reduce the number of identified proteins. RAB and that mining through these neighbors will likely be helpful Finally, we believe that expanding the repertoire of APEX2:RAB to for generating new hypotheses on RAB functions. However, because each cell compartment will help toward identifying specific neigh- of the large number of identified proteins, it will be important to bors as already performed for phosphatases [49], and as observed combine multiple approaches or datasets to help in establishing herein when comparing APEX2:RAB7 to the three early endosomal priorities. Here, the combination of GFP-Trap and APEX2 led us to RABs (Fig 2). identify a novel link between RAB21 and sorting of a subset of Since RAB21 proximity labeling enriched WASH and retromer clathrin-independent cargos. complex members, the functional role between RAB21 and the Given the vast amount of proteins recovered by APEX2:RAB, it is WASH/retromer complexes was further investigated. Importantly, important to use appropriate controls to identify biologically rele- results from the co-immunoprecipitation and pull-down experiments vant candidates. In our hands, normalization to APEX2 and the showed that RAB21 interacts with multiple endogenous WASH and

ª 2019 The Authors EMBO reports 20: 47192 | 2019 13 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

retromer subunits. These findings strongly argue in favor of RAB21 “direct CIE cargos” [86]. This may also explain why RAB21 was also binding to the WASH and retromer complexes. The fact that required for CD44 sorting, even though it did not interact with FAM21, Strumpellin, and CAPZa were observed in pull-down stud- CD44. Given our findings that RAB21 and the WASH/retromer ies further suggests that RAB21 directly interacts with the full WASH complexes are required for “direct CIE cargos” sorting, it could be complex [79]. Nevertheless, the data do not allow to firmly suggested that these sorting events require F-actin generation, most conclude on this latter aspect and more defined structure–function probably for enrichment of cargos in tubules as previously reported studies will be required to map the WASH interaction site, as is also for ß2AR [30] or for the formation of the tubules. Since loss of the case for VPS35 pull-down. It will be furthermore important in RAB21 had a stronger effect on CIE cargo tubule sorting, it is also future studies to define which WASH- or retromer-specific subunit possible that RAB21 plays a structural role by recruiting other effec- (s) is/are bound by RAB21 and to map the binding domains. Finally, tors involved in the generation of these tubules, as also observed for congruent with our co-immunoprecipitation and pull-down results, RAB22 and RAB35 [74,87]. RAB21 depletion partially impaired endosomal recruitment of the Of the various cargos tested herein, only “direct CIE cargos” were WASH and retromer complexes and also led to a decrease in endo- strongly impaired along with MCT1, a recently identified VARP- somal F-actin and cargo trafficking impairments (Fig 8). dependent cargo [20]. Our data thus suggest that MCT1 might also Given the observed role of RAB21 for WASH/retromer functions, represent a “direct CIE cargo”, a possibility worth testing. The one could have expected broader defects in RAB21 KO cells. No requirement of RAB21 for MCT1 trafficking also further strengthens endosomal collapse, as observed in WASH or FAM21-depleted cells our working model in which RAB21 would be involved in recruiting [80], or ectopic endosomal tubulation was observed in RAB21-defi- specific WASH/retromer subcomplexes to direct endosome-to- cient cells [19,28]. This is probably due to the fact that the WASH plasma membrane cargo trafficking. These “specific” WASH/ and retromer complexes are recruited at endosomes through multi- retromer subcomplexes would most probably include VARP while ple pathways. Both SNX3 and RAB7 have been shown to mediate excluding TBC1D5 since VARP and TBC1D5 binding sites on VPS29 endosomal retromer recruitment [13,14,81], while FAM21 binding share the same interface and are thought to be mutually exclusive to VPS35 has been found to be an important determinant of WASH [19,32]. endosomal recruitment [19,25,82]. However, recent studies have Another noteworthy finding from the current proteomic experi- found that WASH can be recruited to endosomes independently of ments was the specific presence of TBC1D23 in APEX2:RAB21 prox- VPS35 [33,83]. Since only a partial reduction in retromer/WASH imity labeling. Recent work has identified TBC1D23 as a bridge endosomal association was observed herein in RAB21 KO cells, a factor for endosomal-to-Golgi carriers [88]. In this study, TBC1D23 model in which RAB21 would be required for endosomal recruit- was shown to interact directly and simultaneously with either ment of a specific (or several) WASH/retromer complex(es) golgin-97 or golgin-245 and the WASH complex [88]. Importantly, involved in the sorting of a subset of cargos is more in accordance TBC1D23, like the WASH complex, was required for TGN46 endo- with the observed data. This latter view is in agreement with newer some-to-Golgi trafficking. It was also proposed that another factor models indicating heterogeneity in retromer sorting decisions and could act together with FAM21 to increase the specificity of captured complexes [20,22,23]. Hence, while necessitating further confirma- vesicles. Given our demonstrated interaction with the WASH tion, an appealing hypothesis would be that RAB21 could act in complex and the presence of TBC1D23 in APEX2:RAB21, RAB21 addition to RAB7 and SNX3, and directly recruit the WASH and could represent such a factor. However, we did not observe strong retromer complexes to specific cargos. Another interesting possibil- defects in either CI-MPR or TGN46 trafficking in RAB21 KO cells. ity would be that RAB21 could function with either SNX3 or RAB7 While this latter observation argues against RAB21 playing a in a co-incidence detection mechanism in order to recruit WASH/ predominant role in this retrieval pathway, RAB21 could nonethe- retromer complexes at endosomes, again to a subset of cargos. In less still be involved in specifying vesicle subtypes that are different this model, RAB21 would specify the type of WASH/retromer from CI-MPR- and TGN46-containing vesicles. subcomplexes involved. The WASH complex has also been recently associated with the Another important aspect of this study is the decrease in detect- retriever complex [33]. This VPS29-, C16orf62-, DSCR3-containing able endosomal F-actin observed in RAB21-depleted cells. The exact core complex, akin to the VPS35, VPS29 and VPS26 cargo retromer mechanism by which RAB21 regulates WASH actin generation is complex, functions with SNX17 to regulate the sorting of a large unclear, given that WASH endosomal localization was only mildly array of cargos. Importantly, proteomic analyses showed distinct affected in RAB21 KO cells. To draw an analogy with the WAVE and overlapping cargos with the retromer [33]. In the present study, complex [84,85], one could speculate that the WASH complex also C16orf62 was enriched in RAB21 proximity labeling whereas DSCR3 requires the action of multiple inputs to drive full activation and and SNX17 were not, while CCC complex members were observed actin polymerization. Ubiquitination was shown to regulate WASH with variable enrichment ratios (Dataset EV2). Hence, RAB21 could activity [70], and it is conceivable that RAB21 could modulate this possibly be involved in regulating WASH and retriever complex process, given its proximity to USP7 (Fig EV4B). Alternatively, it is formation and/or sorting. Our experimental validation of RAB21 also possible that RAB21 could modulate endosomal F-actin inde- was specifically focused on the shared roles of RAB21 with the pendently of WASH, through recruitment of another RAB21 interact- WASH and retromer complexes. In the present instance, retromer- ing protein. Although outside the scope of this study, it would be versus retriever-specific functions were not discriminated due to the interesting to test whether RAB21 can modulate WASH activation use of a VPS29 knockout. Given the role of VPS29 with the retriever directly or indirectly. In keeping with the role of RAB21 in the endo- complex, certain observed phenotypes could have been caused by a somal tubule sorting of SLC3A2 and Basigin, it is worth noting that loss of retriever activity. However, our demonstration that RAB21 F-actin was shown to be required for formation and sorting of functionally interacts with the WASH and retromer complexes is

14 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

robust given that all knockouts shared similar SLC3A2 and Basigin Generation of DNA constructs trafficking impairments. Significantly, loss of VARP function, which is independent of retriever activity and associated with the retromer, GFP:RAB21 variants were PCR-amplified from previously published shared the same RAB21 phenotypes on cargos. Importantly, endoso- pcDNA3 variants [42] and ligated using the In-Fusion HD kit (Clon- mal c16orf62 localization was not decreased by RAB21 deletion, but tech) into a PCR-amplified modified pGLAP1 vector [43] where GFP rather increased (Fig EV5J), thus suggesting that RAB21 could was placed N-terminally to the fusion protein. APEX2:RAB21-WT potentially modulate the balance between retromer and retriever was generated by PCR amplification of RAB21-WT from pCDNA3- complex association with endosomes, possibly through RAB21 inter- GFP:RAB21-WT and subcloned by In-Fusion HD in pGLAP1-APEX2. action with the WASH complex or through VARP. Again, this repre- This vector is a modified version of N-terminal pGLAP1 vector, sents an exciting new possibility which will require further where the GFP was replaced with Myc:APEX2 [90]. APEX2:RAB4a, validation. APEX2:RAB5a, and APEX2:RAB7a were generated by PCR amplifi- In summary, through the use of unbiased proteomics, the present cation of RAB4a, 5a, and 7a from HeLa cell cDNA. cDNA was gener- study allowed uncovering novel functional roles for RAB21 in direct ated with the SuperScript III First-Strand Synthesis kit (Thermo clathrin-independent sorting events. Results also demonstrate the Fisher Scientific) and PCR performed with the Phusion enzyme robustness and ease of APEX2-mediated proximity labeling and its (New England Biolabs). PCR fragments were cloned by In-Fusion applicability to efficiently identify novel RAB-specific functions. This HD in pGLAP1-APEX2. All clones were subsequently sequenced and approach may be further extended to temporally generate dynamic validated. findings on RAB GTPases under various cellular stimulations, thereby leading to a better definition of RAB GTPase regulation and Immunoprecipitations, mass spectrometry, pull-downs, their association with their respective effectors. and immunoblots

A total of 8 × 106 cells were plated in 150-mm plates and grown for Materials and Methods 2 days in appropriate SILAC medium, followed by a 24-h doxycy- cline induction (10 ng/ml). Cell lysis was performed in 2.2 ml of Cell culture CoIP buffer (25 mM Tris–HCl pH 7.4, 1 mM EDTA, 0.1 mM EGTA,

15 mM MgCl2, 150 mM NaCl, 2 mM Na3VO4, 10% glycerol, 1% HCT116 and HeLa Flp-in/T-REx cells and HeLaM cells (gift from T. IGEPAL CA-630, 2× protease inhibitors) per plate, and cells were Yoshimori) were grown in Dulbecco’s modified Eagle’s medium incubated for 20 min on ice. Lysates were cleared by centrifugation (DMEM) supplemented with 10% fetal bovine serum (Wisent) at at 16,000 g for 10 min at 4°C. Cell lysates were quantified through a

37°C and 5% CO2. FT cells were maintained under selection with BSA assay (Pierce), and equal amounts were immunoprecipitated in 100 lg/ml Zeocin and 5 lg/ml blasticidin. FT control cells are the individual tubes with 20 ll of GFP-Trap beads (ChromoTek) and non-recombined parental cells and were used as control in all incubated on a rotator for 2.5 h at 4°C. Beads were washed twice SILAC experiments, as performed previously [43,89]. All cells were with full lysis buffer and twice again with lysis buffer lacking routinely screened for mycoplasma contaminations, and experi- IGEPAL CA-630. This was followed by five washes of 20 mM ments were excluded if contamination was observed. To recombine NH4HCO3. Following these washes, GFP-Trap beads from three inde- the various GFP:RAB21 variants, APEX2:RAB4a, 5a, 7a, and pendent immunoprecipitations (light, medium, and heavy) were APEX2:RAB21 in cells, FT cells were transfected with jetPRIME mixed equally and processed for on-beads digestion and mass spec- (Polyplus), according to manufacturer’s instruction. A RAB/pOG44 trometry analysis following these steps. Proteins were reduced DNA ratio of 1–10 was used, such that 0.1 lg of pGLAP:RAB was 30 min with 10 mM DTT and alkylated 1 h with 15 mM iodoac- cotransfected along with 0.9 lg of pOG44 (Thermo Fisher) in indi- etamide. After iodoacetamide quenching with 15 mM DTT, proteins vidual wells of a 6-well plate. Twenty-four hours following trans- were digested overnight with 1 lg trypsin. Digestion was stopped by fection, recombinant cells were selected with 5 lg/ml blasticidin acidification with 1% formic acid, supernatant was collected, and and 100 lg/ml hygromycin (150 lg/ml for HeLa cells). Following residual peptides were eluted with 60% acetonitrile and 0.1% formic selection, cells were pooled and a polyclonal population was acid. Samples were then dried, resuspended in 0.1% TFA solution, expanded and used for all subsequent experiments. RAB21, RAB4, and desalted on a Zip Tip. Trypsin-digested peptides were loaded RAB5, RAB7 expression was achieved through addition of doxycy- and separated using an Ultimate 3000 nanoLC (Thermo Fisher Scien- cline at 10 ng/ml final concentration for 24 h prior to all experi- tific Inc). Ten microliters of the sample (2 lg) resuspended in 1% ments. (v/v) formic acid was loaded onto a trap column (Acclaim For SILAC experiments, FT cells were maintained in light DMEM PepMap100 C18 column, 0.3 mm id × 5 mm, Dionex Corporation, (R0K0: Arg0, Sigma, A5006; Lys0, Sigma, L5501), while GFP:RAB21 Sunnyvale, CA), and the peptides were separated by a PepMap C18 variants were grown in medium (R6K4: Arg6, Cambridge Isotope nanocolumn (75 lm × 50 cm, Dionex Corporation) with a linear Lab (CIL), CNM-2265; Lys4, CIL, DLM-2640) or heavy (R10K8: gradient of 5–35% solvent B (90% acetonitrile with 0.1% formic Arg10, CIL, CNLM-539; Lys8, CIL, CNLM-291) DMEM depleted of acid) over a 4 h gradient with a constant flow of 200 nl/min. Acqui- arginine and lysine (Life Technologies A14431-01), and supple- sition of the full-scan MS survey spectra (m/z 350–1,600) in profile mented with 10% dialyzed fetal bovine serum (FBS; Invitrogen, mode was performed in a Q Exactive Orbitrap (Thermo Fisher Scien- 26400-044), called SILAC medium. Cells were passaged 5 times to tific Inc) at a resolution of 70,000 using 1,000,000 ions. All unas- ensure proper incorporation of all heavy amino acids, prior to SILAC signed charge states as well as singly, 7 and 8 charged species for experiments. the precursor ions were rejected. To improve the mass accuracy of

ª 2019 The Authors EMBO reports 20: 47192 | 2019 15 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

survey scans, the lock mass option was enabled. Data acquisition 4°C on a rotating wheel. Beads were then washed twice in CoIP was performed using Xcalibur version 2.2 SP1.48. buffer containing 1% IGEPAL CA-630, twice in CoIP buffer minus

GST and GST:RAB21 were purified exactly as described in Ref. IGEPAL CA-630 and five times with 20 mM NH4HCO3. Lysates were [91] for GST:APPL1, and pull-downs were carried out with equal analyzed by immunoblots, and affinity-purified biotinylated proteins amounts of GST or GST:RAB21. Pull-downs were performed as were processed for Western blots or for on-beads digestion and follows. 3.5 × 106 HeLa cells were plated in 100 mm plates, and mass spectrometry analysis as described above. 24 h later, cells were lysed for 20 min on ice with 1 ml of MLB modified buffer (25 mM HEPES, 150 mM NaCl, 1% IGEPAL CA- Experimental design and statistics

630, 10% glycerol, 20 mM MgCl2, 1 mM sodium orthovanadate, 100 lM EGTA and 100 lM GTP) supplemented with protease inhi- For SILAC experiments, two biological repeats were performed in bitors. Lysates were cleared by centrifugation at 16,000 g for 10 min each cell line for each RAB21 variant, yielding a total of four inde- at 4°C. In the meantime, GST and GST:RAB21 sepharose beads were pendent samples per variant analyzed when HeLa and HCT116 data washed three times in MLB modified buffer minus IGEPAL CA-630 were combined. For APEX2 experiments, results were obtained from and incubated for 20 min at 4°C with rotation in modified MLB three independent biological repeats. Controls (FT) were included in buffer without IGEPAL CA-630. Beads were then washed three times all SILAC mass spectrometry experiments. Hence, a total of six inde- with MLB modified buffer and incubated with 900 ll of lysates for pendent FT controls were used in SILAC experiments and three 1 h at 4°C with rotation. Finally, lysates were discarded and beads APEX2-alone controls in APEX2 experiments. Statistical analyses of were washed three times with MLB modified buffer containing objects count or colocalization results were performed using Prism 0.2% IGEPAL CA-630. Media were removed, and 30 llof2× SDS 7 software. Endosomal tubules were assessed manually. Briefly, the loading buffer was added to each sample. number of cells per field was established using DAPI staining, and For immunoblots, protein extracts were separated on 4–20% cells harboring ≥ 2 tubules were counted as positive. An average TGX precast gels (Bio-Rad) and transferred on PVDF (Millipore) per field was established, and multiple fields of at least two indepen- membranes. Antibodies used for immunoblotting were anti-GFP dent experiments were pooled and analyzed using Prism software. (1:500, Roche #11814460001), anti-RAB21 (1:1,000, Sigma #R4405 For all analysis, the average number of object puncta or tubules or or 1:1,000, Invitrogen #PA5-34404), anti-RAB4 (1:1,000, Cell Signal- the average Pearson correlation per cell was tested for normality ing #2167), anti-RAB5 (1:1,000, Cell Signaling #3547), anti-RAB7 using a D’Agostino–Pearson omnibus normality test. Samples that (1:1,000, Cell Signaling #9367), anti-GAPDH-HRP (1:1,000, Cell showed normal distribution were analyzed through unpaired t-tests, Signaling #8884), anti-Myc (1:1,000, Cell Signaling #2278), anti-HA while samples showing a non-normal distribution were compared (1:1,000, Cell Signaling #3724), anti-SLC3A2 (1:800, Cell Signaling through nonparametric Mann–Whitney tests to assess the signifi- #13180), anti-Strumpellin (1:500, Santa Cruz #377146), anti-VPS35 cance between the various conditions. All graphs display SEMs to (1:500, Santa Cruz #374372), anti-VPS26 (1:500, Santa Cruz assess variations within each group. For Western blot quan- #390304), anti-VPS29 (1:500, Santa Cruz #398874), anti-FAM21 and tifications, bands were quantified on the Image Lab software (Bio- anti-WASH (1:10,000, gift from D. Billadeau), anti-CAPZa (1:500, Rad) and normalized to parental cells. One-sample t-tests were Santa Cruz #374302), anti-APPL1 (1:1,000, Cell Signaling #3858), performed for statistical analyses. anti-VARP (1:500, Bethyl Laboratories #A302-997A), anti-tubulin (1:2,500, Sigma #T9026), streptavidin-HRP (1:1,000, Thermo Fisher Mass spectrometry hit selection and network generation #N100), and anti-rabbit and mouse HRP (1:10,000, Jackson Labora- tories #115-035-144 and #115-035-146, respectively). Luminata Forte After analysis by nanoLC-MS/MS, peptides and proteins were identi- (Millipore) or Clarity Max chemiluminescent substrates were used, fied by MaxQuant version 1.5.2.8 using UniProt (Homo sapiens, 16/ and membranes imaged on a Bio-Rad Chemidoc XR station. 07/2013, 88354 entries, Datasets EV1 and EV2). Trypsin/P was the set enzyme, with no cleavages on arginines or lysines preceding a APEX2 proline, with a maximum of two miscleavages being allowed. Mass tolerance of 7 and 20 ppm was used for precursor and fragment A total of 8 × 106 HeLa cells were plated in 150 mm plates and ions, respectively. Fixed carbamidomethyl modification on cysteine induced 24 h later with 10 ng/ml doxycycline for 24 h. Freshly and variable oxidation on methionine and N-terminal acetylation prepared biotin–phenol was added to full DMEM to yield a final were settled. For reliable identification, all proteins needed to concentration of 500 lM. After renewal of the culture medium, cells complete a false discovery rate (FDR) inferior to 1%: All proteins were incubated in DMEM—biotin–phenol—for 30 min at 37°C. whose hits from the forward database were not 100-fold superior to

Biotinylation was achieved by adding freshly prepared H2O2 at a hits from the reverse database were discarded. For SILAC condi- final concentration of 2 mM for exactly 1 min. Reactions were tions, the re-quantification option was selected. A minimum of two halted by removing the media, transferring the cells on ice, and quantified peptides was settled for proteins to be considered. For a performing five washes, each wash for 1 min, in freshly prepared peptide, following its identification, MaxQuant used the median quencher buffer [1× PBS containing 10 mM sodium azide, 10 mM ratio to approximate a SILAC value for that particular peptide. Like- sodium ascorbate, and 5 mM Trolox (Sigma)]. Cells were then wise, in order to establish a protein–SILAC ratio, MaxQuant used processed as detailed for co-immunoprecipitations (CoIPs), with the the median ratio of all identified peptides, a method shown to be following changes. Protein lysates were incubated with 20 llof appropriate for estimation of SILAC ratios [92]. To determine the streptavidin agarose beads (GE Life Sciences) or 20 ll of biotin anti- RAB21 interactome from SILAC GFP-Trap experiments, only body agarose beads (ImmuneChem Pharmaceuticals) for 2.5 h at proteins with an intensity (RAB21-expressed condition)/intensity

16 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

(FT control condition) ratio > 2 in both experimental repeats were #NBP1-49643SS), anti-Glut1 (1:500, Abcam #115730), anti-MCT1 considered as potential RAB21 interactors [43,89]. In the case of (1:100, Genetex #GTX631643), and goat anti-mouse or rabbit Alexa APEX2 affinity purification–mass spectrometry (AP-MS) experi- 488 or 546 (1:250, Thermo Fisher #A11030, A11035, A11029, and ments, the ProHits software suite [93] was used to compare all A11034). DAPI was used to unbiasedly identify fields where images samples (APEX2, APEX2:RAB4a, APEX2:RAB5a, APEX2:RAB7a, were acquired. APEX2:RAB21, and APEX2:2xFYVE (the raw MS data for APEX2:2x- For transferrin uptake, cells were plated at a density of 40,000 FYVE were downloaded from the ProteomeXchange Consortium). cells on (#1.5) glass coverslips and grown for 24 h. Cells were First, SAINTexpress was run on the Crapome platform to establish serum-starved for 30 min in DMEM. Alexa 647-labeled transferrin high probability interactors. Default settings were used to compare was added at 25 lg/ml and allowed to bind to the receptor for user control, in this instance, APEX2 to each APEX2:RABs [52]. 15 min on ice. Cells were washed with cold PBS and incubated in Only proteins with SAINT scores above 0.95 were selected. There- FBS-supplemented DMEM for 0–60 min at 37°C. Cells were washed after, SAINT raw results were extracted and used to generate dot in PBS and fixed with 4% PFA for 15 min at room temperature and plots or cluster plots on a ProHits-viz platform. In order to account mounted for imaging. SLC3A2 (clone MEM-108, Biolegend), Basigin for experimental variations, all samples were normalized on (clone HIM6, Biolegend), and CD44 (clone BJ18, Biolegend) uptake ProHits-viz for dot plot representations. Once RAB21 potential inter- assays were performed exactly as previously described [74]. actors were identified, RAB21 protein interaction networks were Figures 1D, 4B, EV1C, EV4G, and EV5F were acquired on an generated on Cytoscape. Results were combined by cell lines (HeLa Olympus FV1000 confocal microscope equipped with a 60×/1.42NA or HCT116), and once interaction networks were created, proteins Plan Apo N objective at a 1× zoom (colocalization) or a 40×/1.3NA were sorted based on their Genecard and UniProt subcellular local- UPLANFLN objective (LC3 counts). Single Z-sections were acquired, ization data. Each color corresponds to a different localization. and acquisition settings were set to avoid pixel saturation to ensure Cytoscape platform was used to generate biological Reactome proper colocalization and intensity quantification. Images were process enrichments using Reactome plugging among specific gener- adjusted similarly between conditions to accurately represent the ated networks. To compare differential protein enrichments depend- raw data. Figures 3E, 5C, F, and I, 6A, D, and G, and 7B, E, and F, ing on RAB21 status, SILAC RAB21-variant mean ratios were EV5A, D, and J were acquired on a ZEISS LSM880 equipped with a transformed in log2(x) and 0 was imputed to non-available data. 40×/1.4NA Plan Apo objective at a 1.6× zoom. For colocalization Default settings from the Perseus software [94] were used to gener- experiments, single Z-sections were acquired and gain/offset ate hierarchical clustered heat maps. settings were similarly adjusted between conditions. For some anti- bodies, offset was utilized in order to remove non-specific intracel- Immunofluorescence, colocalization, and transferrin or lular background signal and to facilitate endosomal localization antibody uptake visualization. For endosomal tubule imaging, three Z-sections were acquired at 0.36 lm each and maximum intensity projections were A total of 70,000 HeLa or HCT116 cells were plated on glass cover- generated. These maximum intensity projections were used for data slips (#1.5) and grown for 24 h followed by GFP:RAB21 variant quantification and representation. Finally, only linear modifications induction as performed above. Cells were then washed twice with 1× of levels were performed on Photoshop CC2017. Cropping and reso- PBS and fixed for 15 min at room temperature in 4% paraformalde- lution changes to 300 dpi were also performed on Photoshop for hyde in PBS or, depending on the antibody, in 100% methanol at data presentation. Image colocalization was achieved with CellPro- À20°C for 20 min. Cells were then washed three times with PBS for filer [95]. Briefly, GFP:RAB21 cells were identified manually or 5 min each, blocked, and permeabilized for 60 min at room tempera- through a primary object identification module and the correlation ture in PBS containing 5% goat serum and 0.3% Triton X-100. Cells coefficient (Pearson) was measured on original images using the were then incubated in primary antibodies overnight at 4°C in PBS measure correlation module. Number of object puncta per cells was containing 1% BSA and 0.3% Triton X-100. Primary antibodies were also achieved using CellProfiler. For both colocalization and objects washed three times for 5 min in PBS at room temperature and incu- count, all measurements were exported to Prism for statistical bated at room temperature for 1 h in secondary antibodies diluted in analyses. antibody dilution buffer. Following secondary antibody incubation, three 5-min washes in PBS at room temperature were performed and Generation of knockout HeLa cells cells were mounted in DAPI-containing mounting media (Sigma) and subsequently imaged. Antibodies used were anti-EEA1 (1:100, RAB21 knockout HeLa cell populations were engineered using the Cell Signaling #3288; or 1:1,000 BD Biosciences #610456), anti- CRISPR/Cas9 technique. Briefly, guide RNAs were selected based on SLC3A2 (1:800, Cell Signaling #13180; or 1:400 Santa Cruz Ref. [96]. Two independent gRNAs (gRNA-2 “AGTAAATTGGACC- #376815), anti-transferrin receptor (1:100, Cell Signaling #13113), CAATGCA” and gRNA-3 “CCACCTTGAACGAGTAGGCT”) were anti-RAB7 (1:250, Santa Cruz #376362), anti-LAMP1 (1:250, Santa selected and cloned into pSPCas9(BB)-2A-Puro. A total of 1 × 106 Cruz #20011), anti-LC3B (1:100, Cell Signaling #3868), anti-APPL1 HeLa cells were plated in 100-mm dishes and transfected with indi- (1:100, Cell Signaling #3858), streptavidin–Alexa 647 (1:500, vidual plasmids. After 24 h, transfected cells were selected with Thermo Fisher #S21374), anti-FAM21 or anti-WASH (1:1,000, gift 1.5 lg/ml puromycin for 5 days, with media renewed every 48 h. from D. Billadeau), anti-VPS26 (1:100, Abcam #23892), anti-VPS35 Selected cells were amplified and frozen at low passages. Deletion (Santa Cruz, 1:100 #374372), anti-SNX1 (1:100, Abcam #ab995), of the RAB21 gene was verified by PCR and sequencing. Briefly, phalloidin–Alexa 488 (1:1,000, Invitrogen #A12379), anti-CI-MPR genomic DNA isolated from each gRNA-selected cell population was (1:100, Bio-Rad #MCA2048T), anti-TGN46 (1:100, Novus Biological isolated. Amplicons encompassing predicted cut sites were

ª 2019 The Authors EMBO reports 20: 47192 | 2019 17 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

amplified by PCR (using Phusion polymerase) and ligated into comments during the course of this study. We thank the proteomic platform pBluescript using T4 DNA ligase. Individual bacterial clones were at the Université de Sherbrooke for proteomic services and the Photonic selected and sequenced, and the representation of the cutting events microscopy platform for confocal use. Steve Jean and François-Michel Bois- is depicted in Fig EV4E. All experiments performed on knockout vert are members of the FRQS-Funded Centre de Recherche du CHUS. Steve cells were performed at low passages to avoid competition from Jean is a recipient of a Research Chair from the Centre de recherche médicale potential non-targeted wild-type HeLa cells. de l’Université de Sherbrooke (CRMUS). This research was supported by oper- HeLa FAM21, VPS29, and VARP knockout cell populations were ating grants from the Cancer Research Society (CRS), the Natural Sciences performed as in Ref. [97]. Briefly, three independent gRNAs cloned and Engineering Research Council of Canada (NSERC), and the Canadian into the PX330A plasmid were cotransfected with the pEGFP-Puro Institutes of Health Research (CIHR) and by junior faculty salary awards plasmid (addgene #45561) at a ratio of 1:1:1:0.25. Transfections from Canadian Institutes of Health Research (CIHR) and Fonds de Recherche were performed as detailed above. Twenty-four hours after the du Québec—Santé (FRQS) to S.J. transfections, cells were selected for 24 h with 2 lg/ml puromycin. Cells were amplified and used at low passages for all experiments. Author contributions gRNA sequences for VPS29 and FAM21 have been published in Ref. TDO, AL, RL, ML, DJ, LL, and SJ performed experiments. CN quantified colocal- [97], while gRNAs for VARP were AAGAGCCACTCACGTCCTCG, ization data. F-MB assisted and helped with all mass spectrometry analyses. CAGCACGGCTACTGACTATG, and TAAATACGTGGGGACCATGG. FS provided multiple reagents and suggestions. SJ and TDO designed experi- Knockout efficiencies were monitored by Western blot. ments. SJ supervised students and wrote the manuscript.

FACS analysis Conflict of interest The authors declare that they have no conflict of interest. A total of 1 × 106 HeLa parental and gRNA KO 2 and 3 cells were trypsinized, collected, and washed with cold PBS. Cells were fixed 15 min with PFA 4% and permeabilized with methanol 90%, for References 30 min. Cells were labeled (or not) with SLC3A2 rabbit antibody (dilution 1:800, Cell Signaling #13180) in PBS/BSA 0.5% incuba- 1. Stenmark H (2009) Rab GTPases as coordinators of vesicle traffic. Nat tion buffer for one hour. After two PBS washes, cells were incu- Rev Mol Cell Biol 10: 513 – 525 bated for 30 min in secondary Alexa Fluor 488 antibody (dilution 2. Nicot A-S, Laporte J (2008) Endosomal phosphoinositides and human 1:250) diluted in incubation buffer. For each condition, Alexa Fluor diseases. Traffic 9: 1240 – 1249 488 signal intensity was measured with a BD LSR Fortessa cytome- 3. Amoasii L, Hnia K, Laporte J (2012) Myotubularin phosphoinositide phos- ter. Acquisition was made using BD FACSDiva software, and graph- phatases in human diseases. Curr Top Microbiol Immunol 362: 209 – 233 ics and analysis using Flowing software 2.5.1. Mean 488 intensity 4. Santiago-Tirado FH, Bretscher A (2011) Membrane-trafficking sorting signal ratios from SLC3A2 labeled cells versus non-labeled cells hubs: cooperation between PI4P and small GTPases at the trans-Golgi were used to compare SLC3A2 expression levels between the dif- network. Trends Cell Biol 21: 515 – 525 ferent cell lines. Signal intensity distributions from each cell line 5. Jean S, Kiger AA (2012) Coordination between RAB GTPase and phospho- were overlapped on the same frequency histogram. In order to opti- inositide regulation and functions. Nat Rev Mol Cell Biol 13: 463 – 470 mize visualization, data were submitted to virtual gain, such that 6. Rojas AM, Fuentes G, Rausell A, Valencia A (2012) Evolution: the ras f(x) = 60x + 15,000, with x being the intensity signal. The protein superfamily: evolutionary tree and role of conserved amino displayed diagram (Fig EV5H) reflects one representative experi- acids. J Cell Biol 196: 189 – 201 ment from the three performed repeats and represents 10,000 7. Hutagalung AH, Novick PJ (2011) Role of Rab GTPases in membrane counted cells per cell line. traffic and cell physiology. Physiol Rev 91: 119 – 149 8. Barr F, Lambright DG (2010) Rab GEFs and GAPs. Curr Opin Cell Biol 22: 461 – 470 Data availability 9. Itoh T, Satoh M, Kanno E, Fukuda M (2006) Screening for target Rabs of TBC (Tre-2/Bub2/Cdc16) domain-containing proteins based on their Rab- The SILAC and APEX2 mass spectrometry proteomic data from this binding activity. Genes Cells 11: 1023 – 1037 publication have been deposited to the ProteomeXchange Consor- 10. Barr F, Lambright DG (2010) Rab GEFs and GAPs. Curr Opin Cell Biol 22: tium via the PRIDE partner repository (https://www.ebi.ac.uk/ 461 – 470 pride/archive/) and assigned the dataset identifier PXD010950. 11. Barr FA (2013) Rab GTPases and membrane identity: causal or inconse- quential? J Cell Biol 202: 191 – 199 Expanded View for this article is available online. 12. Grosshans BL, Ortiz D, Novick P (2006) Rabs and their effectors: achiev- ing specificity in membrane traffic. Proc Natl Acad Sci USA 103: Acknowledgements 11821 – 11827 We thank M.L. Dubois and D. Lévesque for helpful technical assistance with 13. Rojas R, Van Vlijmen T, Mardones GA, Prabhu Y, Rojas AL, Mohammed S, mass spectrometry and SILAC experiments, V. Delcourt for help with Perseus, Heck AJ, Raposo G, Van Der Sluijs P, Bonifacino JS (2008) Regulation of C. Lavoie for generously providing multiple antibodies, M. Catala for help retromer recruitment to endosomes by sequential action of Rab5 and with FACS analysis, D. Billadeau for kindly providing the FAM21 and WASH Rab7. J Cell Biol 183: 513 – 526 antibodies, and Amy Kiger for support in establishing tools for the current 14. Seaman MNJ, Harbour ME, Tattersall D, Read E, Bright N (2009) study. We also thank all members of the Jean laboratory for insightful Membrane recruitment of the cargo-selective retromer subcomplex is

18 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

catalysed by the small GTPase Rab7 and inhibited by the Rab-GAP recruited on to endosomes by direct interaction with retromer, where TBC1D5. J Cell Sci 122: 2371 – 2382 together they function in export to the cell surface. Dev Cell 29: 15. Seaman MNJ, Marcusson EG, Cereghino JL, Emr SD (1997) Endosome to 591 – 606 golgi retrieval of the vacuolar protein sorting receptor, Vps10p, requires 33. McNally KE, Faulkner R, Steinberg F, Gallon M, Ghai R, Pim D, Langton the function of the VPS29, VPS30, and VPS35 gene products. J Cell Biol P, Pearson N, Danson CM, Nägele H et al (2017) Retriever is a multipro- 137: 79 – 92 tein complex for retromer-independent endosomal cargo recycling. Nat 16. Burd C, Cullen PJ (2014) Retromer: a master conductor of endosome Cell Biol 19: 1214 – 1225 sorting. Cold Spring Harb Perspect Biol 6:a016774 34. Christoforidis S, Miaczynska M, Ashman K, Wilm M, Zhao L, Yip SC, 17. Steinberg F, Gallon M, Winfield M, Thomas EC, Bell AJ, Heesom KJ, Waterfield MD, Backer JM, Zerial M (1999) Phosphatidylinositol-3-OH Tavaré JM, Cullen PJ (2013) A global analysis of SNX27–retromer assem- kinases are Rab5 effectors. Nat Cell Biol 1: 249 – 252 bly and cargo specificity reveals a function in glucose and metal ion 35. Christoforidis S, Zerial M (2000) Purification and identification of novel transport. Nat Cell Biol 15: 461 – 471 rab effectors using affinity chromatography. Methods 20: 403 – 410 18. Seaman MNJ (2012) The retromer complex – endosomal protein recy- 36. Fukuda M (2010) How can mammalian Rab small GTPases be compre- cling and beyond. J Cell Sci 125: 4693 – 4702 hensively analyzed?: Development of new tools to comprehensively 19. Harbour ME, Breusegem SYA, Antrobus R, Freeman C, Reid E, Seaman analyze mammalian Rabs in membrane traffic. Histol Histopathol 25: MNJ (2010) The cargo-selective retromer complex is a recruiting hub for 1473 – 1480 protein complexes that regulate endosomal tubule dynamics. J Cell Sci 37. Gillingham AK, Sinka R, Torres IL, Lilley KS, Munro S (2014) Toward a 123: 3703 – 3717 comprehensive map of the effectors of rab GTPases. Dev Cell 31: 20. Mcgough IJ, Steinberg F, Gallon M, Yatsu A, Ohbayashi N, Heesom KJ, 358 – 373 Fukuda M, Cullen PJ (2014) Identification of molecular heterogeneity in 38. Huttlin EL, Bruckner RJ, Paulo JA, Cannon JR, Ting L, Baltier K, Colby G, SNX27-retromer-mediated endosome-to-plasma-membrane recycling. J Gebreab F, Gygi MP, Parzen H et al (2017) Architecture of the human Cell Sci 127: 4940 – 4953 interactome defines protein communities and disease networks. Nature 21. Zhong Q, Watson MJ, Lazar CS, Hounslow AM, Waltho JP, Gill GN (2005) 545: 505 – 509 Determinants of the endosomal localization of sorting nexin 1. Mol Biol 39. Alanko J, Mai A, Jacquemet G, Schauer K, Kaukonen R, Saari M, Goud B, Cell 16: 2049 – 2057 Ivaska J (2015) Integrin endosomal signalling suppresses anoikis. Nat Cell 22. Kvainickas A, Jimenez Orgaz A, Nägele H, Hu Z, Dengjel J, Steinberg F Biol 17: 1412 – 1421 (2017) Cargo-selective SNX-BAR proteins mediate retromer trimer inde- 40. Pellinen T, Tuomi S, Arjonen A, Wolf M, Edgren H, Meyer H, Grosse R, pendent retrograde transport. J Cell Biol 216: 3677 – 3693 Kitzing T, Rantala JK, Kallioniemi O (2008) Integrin trafficking regulated 23. Simonetti B, Danson CM, Heesom KJ, Cullen PJ (2017) Sequence-depen- by Rab21 is necessary for cytokinesis. Dev Cell 15: 371 – 385 dent cargo recognition by SNX-BARs mediates retromer-independent 41. Burgo A, Sotirakis E, Simmler M-C, Verraes A, Chamot C, Simpson JC, transport of CI-MPR. J Cell Biol 216: 3695 – 3712 Lanzetti L, Proux-Gillardeaux VER, Galli T (2009) Role of Varp, a Rab21 24. Mcgough IJ, Cullen PJ (2011) Recent advances in retromer biology. Traf- exchange factor and TI-VAMP/VAMP7 partner, in neurite growth. EMBO fic 12: 963 – 971 Rep 10: 1117 – 1124 25. Helfer E, Harbour ME, Henriot V, Lakisic G, Sousa-Blin C, Volceanov L, 42. Jean S, Cox S, Nassari S, Kiger AA (2015) Starvation-induced MTMR13 Seaman MNJ, Gautreau A (2013) Endosomal recruitment of the WASH and RAB21 activity regulates VAMP8 to promote autophagosome-lyso- complex: active sequences and mutations impairing interaction with the some fusion. EMBO Rep 16: 297 – 311 retromer. Biol Cell 105: 191 – 207 43. Drissi R, Dubois M-L, Douziech M, Boisvert F-M (2015) Quantitative 26. Zavodszky E, Seaman MNJ, Moreau K, Jimenez-Sanchez M, Breusegem proteomics reveals dynamic interactions of the minichromosome main- SY, Harbour ME, Rubinsztein DC (2014) Mutation in VPS35 associated tenance complex (MCM) in the cellular response to etoposide induced with Parkinson’s disease impairs WASH complex association and inhibits DNA damage. Mol Cell Proteomics 14: 2002 – 2013 autophagy. Nat Commun 5: 3828 44. Simpson JC, Griffiths G, Wessling-Resnick M, Fransen JAM, Bennett H, 27. Gomez TS, Billadeau DD (2009) A FAM21-containing WASH complex Jones AT (2004) A role for the small GTPase Rab21 in the early endo- regulates retromer-dependent sorting. Dev Cell 17: 699 – 711 cytic pathway. J Cell Sci 117: 6297 – 6311 28. Derivery E, Sousa C, Gautier JJ, Lombard B, Loew D, Gautreau A (2009) 45. Yuan Q, Ren C, Xu W, Petri B, Zhang J, Zhang Y, Kubes P, Wu D, Tang W The Arp2/3 activator WASH controls the fission of endosomes through a (2017) PKN1 directs polarized RAB21 vesicle trafficking via RPH3A and is large multiprotein complex. Dev Cell 17: 712 – 723 important for neutrophil adhesion and ischemia-reperfusion injury. Cell 29. Seaman MNJ, Gautreau A, Billadeau DD (2013) Retromer-mediated Rep 19: 2586 – 2597 endosomal protein sorting: all WASHed up!. Trends Cell Biol 23: 46. Martell JD, Deerinck TJ, Sancak Y, Poulos TL, Mootha VK, Sosinsky GE, 522 – 528 Ellisman MH, Ting AY (2012) Engineered ascorbate peroxidase as a 30. Temkin P, Lauffer B, Jäger S, Cimermancic P, Krogan NJ, von Zastrow genetically encoded reporter for electron microscopy. Nat Biotechnol 30: M(2011) SNX27 mediates retromer tubule entry and endosome-to- 1143 – 1148 plasma membrane trafficking of signalling receptors. Nat Cell Biol 13: 47. Lee S-Y, Kang M-G, Shin S, Kwak C, Kwon T, Seo JK, Kim J-S, Rhee H- 715 – 721 W(2017) Architecture mapping of the inner mitochondrial membrane 31. Steinberg F, Heesom KJ, Bass MD, Cullen PJ (2012) SNX17 protects inte- proteome by chemical tools in live cells. J Am Chem Soc 139: grins from degradation by sorting between lysosomal and recycling 3651 – 3662 pathways. J Cell Biol 197: 219 – 230 48. Lee S-Y, Kang M-G, Park J-S, Lee G, Ting AY, Rhee H-W (2016) APEX fin- 32. Hesketh GG, Pérez-Dorado I, Jackson LP, Wartosch L, Schäfer IB, Gray SR, gerprinting reveals the subcellular localization of proteins of interest. McCoy AJ, Zeldin OB, Garman EF, Harbour ME et al (2014) VARP is Cell Rep 15: 1837 – 1847

ª 2019 The Authors EMBO reports 20: 47192 | 2019 19 of 21 EMBO reports RAB21 regulates CIE cargo sorting Tomas Del Olmo et al

49. St-Denis N, Gupta GD, Lin Z-Y, Gonzalez-Badillo B, Veri AO, Knight JDR, molecular rheostat to promote WASH-dependent endosomal protein Rajendran D, Couzens AL, Currie KW, Tkach JM et al (2016) Phenotypic recycling and is mutated in a human neurodevelopmental disorder. Mol and interaction profiling of the human phosphatases identifies diverse Cell 59: 956 – 969 mitotic regulators. Cell Rep 17: 2488 – 2501 67. Udeshi ND, Pedram K, Svinkina T, Fereshetian S, Myers SA, Aygun O, 50. Sönnichsen B, De Renzis S, Nielsen E, Rietdorf J, Zerial M (2000) Distinct Krug K, Clauser K, Ryan D, Ast T et al (2017) Antibodies to biotin enable membrane domains on endosomes in the recycling pathway visualized by large-scale detection of biotinylation sites on proteins. Nat Methods 14: multicolor imaging of Rab4, Rab5, and Rab11. J Cell Biol 149: 901 – 914 1167 – 1170 51. De Renzis S, Sönnichsen B, Zerial M (2002) Divalent Rab effectors regu- 68. Pellinen T, Arjonen A, Vuoriluoto K, Kallio K, Fransen JAM, Ivaska J (2006) late the sub-compartmental organization and sorting of early endo- Small GTPase Rab21 regulates cell adhesion and controls endosomal somes. Nat Cell Biol 4: 124 – 133 traffic of beta1-integrins. J Cell Biol 173: 767 – 780 52. Choi H, Larsen B, Lin Z-Y, Breitkreutz A, Mellacheruvu D, Fermin D, Qin 69. Ran FA, Hsu PD, Wright J, Agarwala V, Scott DA, Zhang F (2013) Genome ZS, Tyers M, Gingras A-C, Nesvizhskii AI (2010) SAINT: probabilistic scor- engineering using the CRISPR-Cas9 system. Nat Protoc 8: 2281 – 2308 ing of affinity purification–mass spectrometry data. Nat Methods 8: 70. Hao Y-H, Doyle JM, Ramanathan S, Gomez TS, Jia D, Xu M, Chen ZJ, Bill- 70 – 73 adeau DD, Rosen MK, Potts PR (2013) Regulation of WASH-dependent 53. Knight JDR, Choi H, Gupta GD, Pelletier L, Raught B, Nesvizhskii AI, actin polymerization and protein trafficking by ubiquitination. Cell 152: Gingras A-C (2017) ProHits-viz: a suite of web tools for visualizing inter- 1051 – 1064 action proteomics data. Nat Methods 14: 645 – 646 71. Seaman MNJ (2009) Enhanced snapshot: endosome-to-golgi retrieval. 54. Jean S, Cox S, Schmidt EJ, Robinson FL, Kiger A (2012) Sbf/MTMR13 coor- Cell 139: 1198 .e1 dinates PI(3)P and Rab21 regulation in endocytic control of cellular 72. Maldonado-Báez L, Williamson C, Donaldson JG (2013) Clathrin-indepen- remodeling. Mol Biol Cell 23: 2723 – 2740 dent endocytosis: a cargo-centric view. Exp Cell Res 319: 2759 – 2769 55. De Franceschi N, Hamidi H, Alanko J, Sahgal P, Ivaska J (2015) Integrin 73. Eyster CA, Higginson JD, Huebner R, Porat-Shliom N, Weigert R, Wu WW, traffic – the update. J Cell Sci 128: 839 – 852 Shen R-F, Donaldson JG (2009) Discovery of new cargo proteins that enter 56. Chotard L, Mishra AK, Sylvain M-A, Tuck S, Lambright DG, Rocheleau CE cells through clathrin-independent endocytosis. Traffic 10: 590 – 599 (2010) TBC-2 regulates RAB-5/RAB-7-mediated endosomal trafficking in 74. Maldonado-Báez L, Cole NB, Krämer H, Donaldson JG (2013) Micro- Caenorhabditis elegans. Mol Biol Cell 21: 2285 – 2296 tubule-dependent endosomal sorting of clathrin-independent cargo by 57. Fuchs E, Haas AK, Spooner RA, Yoshimura S-I, Lord JM, Barr FA Hook1. J Cell Biol 201: 233 – 247 (2007) Specific Rab GTPase-activating proteins define the Shiga toxin 75. Kvainickas A, Orgaz AJ, Nägele H, Diedrich B, Heesom KJ, Dengjel J, and epidermal growth factor uptake pathways. JCellBiol177: Cullen PJ, Steinberg F (2017) Retromer- and WASH-dependent sorting of 1133 – 1143 nutrient transporters requires a multivalent interaction network with 58. Lobingier BT, Hüttenhain R, Eichel K, Miller KB, Ting AY, von Zastrow M, ANKRD50. J Cell Sci 130: 382 – 395 Krogan NJ (2017) An approach to spatiotemporally resolve protein inter- 76. Zhang X (2006) Varp is a Rab21 guanine nucleotide exchange factor and action networks in living cells. Cell 169: 350 – 360.e12 regulates endosome dynamics. J Cell Sci 119: 1053 – 1062 59. Priya A, Kalaidzidis IV, Kalaidzidis Y, Lambright D, Datta S (2014) Molec- 77. Simonsen A, Lippé R, Christoforidis S, Gaullier JM, Brech A, Callaghan J, ular insights into Rab7-mediated endosomal recruitment of core Toh BH, Murphy C, Zerial M, Stenmark H (1998) EEA1 links PI(3)K func- retromer: deciphering the role of Vps26 and Vps35. Traffic 16: 68 – 84 tion to Rab5 regulation of endosome fusion. Nature 394: 494 – 498 60. Schindler C, Chen Y, Pu J, Guo X, Bonifacino JS (2015) EARP is a multi- 78. Kim DI, Cutler JA, Na CH, Reckel S, Renuse S, Madugundu AK, Tahir R, subunit tethering complex involved in endocytic recycling. Nat Cell Biol Goldschmidt HL, Reddy KL, Huganir RL et al (2018) BioSITe: a method 17: 639 – 650 for direct detection and quantitation of site-specific biotinylation. J 61. Gershlick DC, Schindler C, Chen Y, Bonifacino JS (2016) TSSC1 is novel Proteome Res 17: 759 – 769 component of the endosomal retrieval machinery. Mol Biol Cell 27: 79. Derivery E, Gautreau A (2010) Assaying WAVE and WASH complex 2867 – 2878 constitutive activities toward the Arp2/3 complex. Methods Enzymol 484: 62. Rottner K, Hänisch J, Campellone KG (2010) WASH, WHAMM and JMY: 677 – 695 regulation of Arp2/3 complex and beyond. Trends Cell Biol 20: 650 – 661 80. Gomez TS, Gorman JA, Artal-Martinez De Narvajas A, Koenig AO, Billa- 63. Shin H-W, Hayashi M, Christoforidis S, Lacas-Gervais S, Hoepfner S, deau DD (2012) Trafficking defects in WASH-knockout fibroblasts origi- Wenk MR, Modregger J, Uttenweiler-Joseph S, Wilm M, Nystuen A nate from collapsed endosomal and lysosomal networks. Mol Biol Cell et al (2005) An enzymatic cascade of Rab5 effectors regulates phos- 23: 3215 – 3228 phoinositide turnover in the endocytic pathway. J Cell Biol 170: 81. Harterink M, Port F, Lorenowicz MJ, Mcgough IJ, Silhankova M, Betist 607 – 618 MC, van Weering JRT, Heesbeen RGHPV, Middelkoop TC, Basler K et al 64. Muhammad E, Levitas A, Singh SR, Braiman A, Ofir R, Etzion S, Sheffield (2011) A SNX3-dependent retromer pathway mediates retrograde trans- VC, Etzion Y, Carrier L, Parvari R (2015) PLEKHM2 mutation leads to port of the Wnt sorting receptor Wntless and is required for Wnt secre- abnormal localization of lysosomes, impaired autophagy flux and asso- tion. Nat Cell Biol 13: 914 – 923 ciates with recessive dilated cardiomyopathy and left ventricular 82. Jia D, Gomez TS, Billadeau DD, Rosen MK (2012) Multiple repeat noncompaction. Hum Mol Genet 24: 7227 – 7240 elements within the FAM21 tail link the WASH actin regulatory complex 65. Shin C-S, Mishra P, Watrous JD, Carelli V, Aurelio MDR, Jain M, Chan DC to the retromer. Mol Biol Cell 23: 2352 – 2361 (2017) The glutamate/cystine xCT antiporter antagonizes glutamine 83. Buckley CM, Gopaldass N, Bosmani C, Johnston SA, Soldati T, Insall RH, metabolism and reduces nutrient flexibility. Nat Commun 8: 1 – 11 King JS (2016) WASH drives early recycling from macropinosomes and 66. Hao Y-H, Fountain MD Jr, Tacer KF, Xia F, Bi W, Kang S-HL, Patel A, phagosomes to maintain surface phagocytic receptors. Proc Natl Acad Rosenfeld JA, Le Caignec C, Isidor B et al (2015) USP7 acts as a Sci USA 113:E5906 – E5915

20 of 21 EMBO reports 20: 47192 | 2019 ª 2019 The Authors Tomas Del Olmo et al RAB21 regulates CIE cargo sorting EMBO reports

84. Koronakis V, Hume PJ, Humphreys D, Liu T, Hørning O, Jensen ON, 92. Guan X, Rastogi N, Parthun MR, Freitas MA (2014) SILAC peptide ratio McGhie EJ (2011) WAVE regulatory complex activation by cooperating calculator: a tool for SILAC quantitation of peptides and post-transla- GTPases Arf and Rac1. Proc Natl Acad Sci USA 108: 14449 – 14454 tional modifications. J Proteome Res 13: 506 – 516 85. Lebensohn AM, Kirschner MW (2009) Activation of the WAVE complex 93. Liu G, Zhang J, Larsen B, Stark C, Breitkreutz A, Lin Z-Y, Breitkreutz B-J, by coincident signals controls actin assembly. Mol Cell 36: 512 – 524 Ding Y, Colwill K, Pasculescu A et al (2010) ProHits: integrated software 86. Donaldson JG, Johnson DL, Dutta D (2016) Rab and Arf G proteins in endo- for mass spectrometry– based interaction proteomics. Nat Biotechnol 28: somal trafficking and cell surface homeostasis. Small GTPases 7: 247 – 251 1015 – 1017 87. Dutta D, Donaldson JG (2015) Sorting of clathrin-independent cargo 94. Tyanova S, Temu T, Sinitcyn P, Carlson A, Hein MY, Geiger T, Mann M, proteins depends on Rab35 delivered by clathrin-mediated endocytosis. Cox J (2016) The Perseus computational platform for comprehensive Traffic 16: 994 – 1009 analysis of (prote)omics data. Nat Methods 13: 731 – 740 88. Shin JJH, Gillingham AK, Begum F, Chadwick J, Munro S (2017) TBC1D23 95. Carpenter AE, Jones TR, Lamprecht MR, Clarke C, Kang IH, Friman O, is a bridging factor for endosomal vesicle capture by golgins at the Guertin DA, Chang JH, Lindquist RA, Moffat J et al (2006) Cell Profiler: trans-Golgi. Nat Cell Biol 19: 1424 – 1432 image analysis software for identifying and quantifying cell phenotypes. 89. Dubois M-L, Bastin C, Lévesque D, Boisvert F-M (2016) Comprehensive Genome Biol 7:R100 characterization of minichromosome maintenance complex (MCM) 96. Sullender M, Hegde M, Vaimberg EW, Donovan KF, Smith I, Tothova Z, protein interactions using affinity and proximity purifications coupled to Wilen C, Orchard R, Virgin HW, Doench JG et al (2016) Optimized sgrNA mass spectrometry. J Proteome Res 15: 2924 – 2934 design to maximize activity and minimize off-target effects of crisPr- 90. Lam SS, Martell JD, Kamer KJ, Deerinck TJ, Ellisman MH, Mootha VK, Ting cas9. Nat Biotechnol 34: 184 – 191 AY (2014) Directed evolution of APEX2 for electron microscopy and prox- 97. Jimenez Orgaz A, Kvainickas A, Nägele H, Denner J, Eimer S, Dengjel J, imity labeling. Nat Methods 12: 51 – 54 Steinberg F (2018) Control of RAB7 activity and localization through the 91. Jean S, Kiger AA (2016) RAB21 activity assay using GST-fused APPL1. Bio retromer-TBC1D5 complex enables RAB7-dependent mitophagy. EMBO J Protoc 6:e1738 37: 235 – 254

ª 2019 The Authors EMBO reports 20: 47192 | 2019 21 of 21 © 2019. Published by The Company of Biologists Ltd | Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336

RESEARCH ARTICLE RAB21 interacts with TMED10 and modulates its localization and abundance Tomas Del Olmo, Camille Lacarriere-Keïta, Caroline Normandin, Dominique Jean, François-Michel Boisvert and Steve Jean*

ABSTRACT are inhibited by GAPs (GTPase-activated proteins) that trigger Membrane trafficking controls vesicular transport of cargo between intrinsic hydrolytic activity of RABs (Barr and Lambright, 2010). cellular compartments. Vesicular trafficking is essential for cellular Once activated, RABs recruit a large number of effectors to achieve homeostasis and dysfunctional trafficking is linked to several their functions (Grosshans et al., 2006). Moreover, RABs can pathologies such as neurodegenerative diseases. Following directly interact with cargo to regulate their trafficking (Pellinen endocytosis, early endosomes act as sorting stations of internalized et al., 2006). materials, routing cargo toward various fates. One important class of RAB21 regulates integrin internalization by binding directly to α β membrane trafficking regulators are RAB GTPases. RAB21 has 5 1 integrin (Pellinen et al., 2006). Initially described as an early been associated with multiple functions and regulates integrin endosomal RAB (Simpson et al., 2004), RAB21 has been shown to internalization, endosomal sorting of specific clathrin-independent be involved in various specific functions. It mediates EGFR cargo and autophagy. Although RAB21 is mostly associated with degradation (Yang et al., 2012), controls neurite extensions (Burgo early endosomes, it has been shown to mediate a specific sorting et al., 2009, 2012), regulates autophagic flux (Jean et al., 2015) and event at the Golgi. From mass spectrometry data, we identified a was recently shown to be associated with clathrin-independent GTP-favored interaction between RAB21 and TMED10 and 9, cargo trafficking via WASH and retromer complexes (Del Olmo essential regulators of COPI and COPII vesicles. Using RAB21 et al., 2019). In the same study, potential interactions between knockout cells, we describe the role of RAB21 in modulating TMED10, TMED9 and RAB21 were observed by quantitative mass TMED10 Golgi localization. Taken together, our study suggests a spectrometry analysis (Del Olmo et al., 2019). new potential function of RAB21 in modulating TMED10 trafficking, with TMED10 and TMED9 both belong to the p24 family of proteins relevance to neurodegenerative disorders. (Pastor-Cantizano et al., 2016). These proteins are mostly localized between the ER and Golgi compartments, cycling between both of KEY WORDS: Membrane trafficking, TMED2, TMED9, p24 family them and mediating cargo transport through COPI and COPII vesicles (Popoff et al., 2011). All proteins of the p24 family are INTRODUCTION composed of a GOLD intraluminal domain allowing interaction Membrane trafficking, which represents all vesicular exchanges with cargo (Anantharaman and Aravind, 2002), a coiled-coil between organelles and cellular compartments, is highly regulated domain, a transmembrane domain mediating homo- and hetero- and essential for cellular homeostasis (Vicinanza et al., 2008). dimerization (Contreras et al., 2012) and a short cytosolic domain Indeed, trafficking defects are involved in a large panel of diseases, involved in protein sorting (Contreras et al., 2004). The cytosolic such as neurological pathologies (Stenmark, 2009). One important domain has been demonstrated to interact with the GTP-bound class of membrane trafficking regulators are the RAB GTPases ARF1 (Gommel et al., 1999), allowing recruitment of ERD2 and (Jean and Kiger, 2012). With almost 70 members, RABs represent formation of COPI vesicles (Majoul et al., 2001). Importantly, the largest family of small GTPases in humans (Rojas et al., 2012). TMED10 and TMED2 expression is necessary to maintain ER and These proteins mediate each step of vesicular trafficking, from Golgi integrity (Montesinos et al., 2012). RAB21 has been shown to membrane budding to vesicle transport, to fusion with target sort VAMP7 at the Golgi (Burgo et al., 2012), and several members organelles (Hutagalung and Novick, 2011). Given their roles in of the p24 family have been identified as potential RAB21-binding trafficking, RABs are tightly regulated (Barr and Lambright, 2010). proteins by mass spectrometry analysis (Del Olmo et al., 2019). Thus, RABs cycle between their active GTP-bound form and Therefore, we characterized the TMED10 and RAB21 interaction inactive GDP-bound form. RABs are activated by GEFs (guanine using biochemical and genetic approaches. This allowed us to exchange factors), which catalyze the exchange of GDP to GTP and define a RAB21 requirement for appropriate TMED10 localization and protein abundance.

Faculte de Me decine et des Sciences de la Sante ,De partement d’anatomie et de RESULTS biologie cellulaire, Universite de Sherbrooke, 3201, Rue Jean Mignault, TMED10 interacts indirectly with RAB21 Sherbrooke, Que bec, Canada J1E 4K8. Although the functional association of TMED10 with ARF1 has *Author for correspondence ([email protected]) been well characterized (D’Souza-Schorey and Chavrier, 2006), potential interactions with RAB family GTPases have only been S.J., 0000-0001-6881-5781 shown by proteomic analysis (Hein et al., 2015) or genetic screens This is an Open Access article distributed under the terms of the Creative Commons Attribution (Blomen et al., 2015), but have not been further assessed. Our recent License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed. mass spectrometry data identified a potential interaction between RAB21, TMED10 and TMED9. Strong enrichment of these two

Received 7 June 2019; Accepted 18 August 2019 proteins with wild-type or the GTP-bound form of RAB21 have Biology Open

1 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019 RESEARCH ARTICLE Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336 been observed using quantitative interactomics experiments (Del identified VARP- and RAB21-dependent functions in VAMP7 Olmo et al., 2019), in both HeLa and HCT116 cells (see fig. S1A trafficking at the Golgi in neuronal cells (Burgo et al., 2012). from Del Olmo et al., 2019). Therefore, we assessed the functional relationship between RAB21 To validate the interaction between RAB21 and TMED10, GFP and TMED10 in RAB21 knockout cells. Phenotype specificity was co-immunoprecipitation assays were performed in Flp-In/T-REx ensured by generating two independent cell populations using two HeLa and HCT116 cell lines that express GFP:RAB21 close to independent guide RNAs. These cells have previously been endogenous levels in response to doxycycline treatment, as validated by sequencing and western blot analyses (Del Olmo described previously (Del Olmo et al., 2019). Consistent with the et al., 2019). proteomics data, endogenous TMED10 was enriched in GFP: Parental and RAB21 knockout cells were transfected with GFP: RAB21 immunoprecipitations in both HeLa (Fig. 1A) and HCT116 TMED10 (Blum and Lepier, 2008) and TMED10 localization at cis- (Fig. 1B) cells, while an unrelated golgi protein, TGN38, was not and trans-Golgi was investigated by colocalization with GM130, (Fig. S1B). To test whether the identified interaction was direct, TGN46 and ci-MPR, respectively (Fig. 3A; Fig. S2A,B). GST-pulldown assays were performed using purified GST:RAB21 Interestingly, RAB21 deletion reduced TMED10 localization in or GST:RAB21-Q78L (GTP-bound) and incubated with HeLa cell the cis-Golgi compartment (Fig. 3A). Multiple TMED10 puncta lysates. Pulldown assays showed no specific enrichment of were observed outside the cis-Golgi in these cells. Moreover, TMED10 or TMED2 with either wild-type or GTP-bound forms, TMED10 colocalization with GM130 significantly decreased in although VPS35 (Del Olmo et al., 2019) was present with GST: both HeLa-RAB21 KO cells compared to parental cells (Fig. 3B). RAB21 (Fig. S1C,D). From these results, we conclude that RAB21 TMED10 colocalization with TGN46 also decreased in both HeLa- interacts indirectly with TMED10. RAB21-KO cell populations, however only the gRNA-2 population showed a statistical difference (Fig. S2C). Surprisingly, no difference TMED10 interacts preferentially with activated RAB21 was observed between TMED10 and ci-MPR colocalization To assess the RAB21 interaction with TMED10, we performed (Fig. S2D). Given that ci-MPR also labels endocytic vesicles, an proximity ligation assays (PLA). HeLa cells were singly or co- increased localization of TMED10:GFP in these vesicles could transfected with TMED10:3xHA and either V5:RAB21-WT, V5: potentially compensate for the observed difference at the Golgi and RAB21-Q78L (GTP-bound) or V5:RAB21-T33N (GDP-bound) yield similar Pearson correlation values. Taken together, these results variants. The number of PLA puncta per cell (indicative of suggest the role of RAB21 in TMED10 maintenance or targeting at TMED10 and RAB21 proximity) was counted through confocal the cis-Golgi compartment and potentially at the trans-Golgi as well. imaging and automated image analysis. While transfection of either TMED10 or RAB21 alone yielded a maximum of nine puncta per RAB21 depletion reduces TMED10 protein levels cell (Fig. 2A,B), co-transfection of TMED10 with RAB21-WT or Given that TMED10 was mis-localized in RAB21 knockout cells, RAB21-Q78L led to a considerable increase in the number of PLA we assessed whether TMED10 protein levels were also affected. puncta per cell, reaching an average of 42 puncta per cell in RAB21- Using western blotting, we compared relative TMED10 expression WT cells. Notably, the number of PLA puncta per cell was in parental and RAB21-KO HeLa cells (Fig. 4A). Relative protein significantly higher in RAB21-WT and RAB21-Q78L variants quantification showed that in both RAB21 knockout populations, compared to RAB21-T33N (Fig. 2A,C). These PLA results are in TMED10 expression was almost twice as low as in the control accordance with the proteomics data and indicate that the interaction (Fig. 4B). Since p24 family members are known to oligomerize between RAB21 and TMED10 is increased upon RAB21 (Contreras et al., 2012) and depletion of TMED10 affects other p24 activation. family members (Pastor-Cantizano et al., 2016), we also analyzed TMED2 protein abundance. In accordance with the result observed RAB21 knockout affects TMED10 localization in cells for TMED10, we found that TMED2 expression was also altered in TMED10 has been reported to localize at the ER-Golgi interface. RAB21 knockout cells (Fig. 4C). We assessed if this was due to On the other hand, RAB21 localizes mostly on early endosomes changes in transcription or in protein stability. Quantitative PCR except for the dominant negative RAB21 (T33N), which is strongly analyses of TMED2 and TMED10 did not highlight any significant associated with the Golgi (Simpson et al., 2004). A previous study difference in expression (Fig. 4D,E). Similarly, a cycloheximide chase did not show apparent differences in stability (Fig. 4F,G). We further assessed if TMED10 half-life was modulated by proteasome or lysosomal degradation. MG132 or Bafilomycin A1 treatments, which block proteasome or lysosome functions, respectively, did not significantly impact TMED10 protein levels in either parental or RAB21 knockout cells (Fig. S3A,B). From these results, we conclude that RAB21 is required for expression of TMED10 and TMED2, that TMED10 has a long half-life in HeLa cells and that the exact mechanism leading to TMED2 and 10 downregulation needs to be elucidated.

DISCUSSION First identified to play a specific role in integrin trafficking (Pellinen Fig. 1. RAB21 interacts with TMED10. (A,B) Western blotting showing et al., 2006), RAB21 is now associated with several other functions GFP-trap immunoprecipitation in HeLa (A) and HCT116 (B) cells. (Del Olmo et al., 2019; Jean et al., 2015; Alanko et al., 2015). In the Endogenous TMED10 was blotted and showed a specific enrichment with GFP:RAB21-WT compared to Flp-In/T-REx (FT) control in both cell lines. present study, we confirmed previous mass spectrometry data showing Lysates represent 2% of input and n=4 independent experiments performed a potential interaction between RAB21 and TMED10 (Del Olmo in each cell line. et al., 2019). We conclude that activated RAB21 interacts indirectly Biology Open

2 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019 RESEARCH ARTICLE Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336

Fig. 2. RAB21 nucleotide status modulates the interaction with TMED10. (A) PLA immunofluorescence showing respective proximities between TMED10:HA and either V5:RAB21 WT, Q78L or T33N variants. TMED10:HA and V5:RAB21 only are controls. PLA puncta are stained in red and nucleus in blue. Dotted lines define individual cells. Scale bars: 10 µm, n=2 independent experiments. (B) Quantification of PLA controls shown in A. Histogram represents average number of PLA puncta per cell, error bars are s.e.m. No statistical analysis was performed. (C) Quantification of PLA experiments shown in A. Histogram represents average number of PLA dots per cell in each RAB21 variant condition, error bars are s.e.m. Mann–Whitney tests were used for statistical analysis. n.s., not significant. with TMED10 as evidenced by immunoprecipitation analysis, preferential interaction of TMED10 with activated RAB21 thus proximity ligation assays and pull-down data. We further show that suggests that TMED10 could act as a RAB21 effector or that RAB21 is required for TMED10 localization in the Golgi and that activated RAB21 could influence the interaction between TMED10 RAB21 influences TMED10 and TMED2 protein expression. and its specific cargo. The data from RAB21 knockout cells TMED10 is mostly observed at the cis-Golgi (Pastor-Cantizano indicates that both possibilities are plausible. However, it is unlikely et al., 2016), compared to RAB21, which is mostly endosomal. that TMED10 would act as a RAB21 effector, due to the lack of a However, RAB21 has been observed at the Golgi in neurons direct interaction between the two proteins. Hence, a Golgi- where it regulates, through VARP, the sorting of VAMP7 (Burgo associated RAB21 pool could retain TMED10 at the Golgi by et al., 2012). Hence, it is possible that another function of VARP interacting with TMED10, possibly through an unknown protein. would be regulation of RAB21 interaction with TMED10. The Alternately, RAB21 could strengthen or weaken TMED10

Fig. 3. TMED10 Golgi localization is altered in RAB21 knockout cells. (A) TMED10:GFP colocalization with endogenous GM130 in parental and RAB21 knockout cells. gRNA-2 and -3 are two independent populations of RAB21 knockout HeLa cells. TMED10:GFP is stained in green, GM130 in red and nucleus in blue. Dotted squares are magnified on the right, arrows show TMED10:GFP-only-labeled vesicles. n=3 independent experiments. Scale bars: 10 µm; 5 µm in zoom. (B) Quantification of TMED10 and GM130 colocalization. Histogram represents average Pearson correlation per cell, error bars are s.e.m. Mann–Whitney tests were used for statistical analysis. Biology Open

3 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019 RESEARCH ARTICLE Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336

Fig. 4. RAB21 knockout reduces TMED10 and TMED2 protein levels. (A) Western blotting showing TMED10 and TMED2 protein levels in parental and RAB21 knockout cells. Endogenous TMED10, TMED2, RAB21 and Tubulin were blotted. Tubulin was used as a housekeeping gene, n=4 independent experiments. (B,C) Quantification of relative protein expression shown in A. TMED10/Tubulin (B) and TMED2/Tubulin (C) protein ratios were normalized to parental cells, error bars are s.e.m. One-sample t-tests were used for statistical analysis. (D,E) Quantitative PCR analysis of TMED10 and TMED2 transcripts, respectively, n=4 independent experiments, error bars are s.e.m. Unpaired t-tests were used for statistical analysis. No statistical differences were observed and are thus not displayed. (F,G) TMED10 stability assay. Cycloheximide chases were performed to monitor TMED10 stability in parental and RAB21 gRNA-3 KO HeLa cells, n=3 independent experiments. (F) Endogenous TMED10, RAB21 and GAPDH were assessed through western blotting. (G) Quantification of relative protein expression shown in F, error bars are s.e.m. Unpaired t-tests were performed and no statistical differences were observed and therefore are not displayed. interaction with cargo, as observed for ARF1 (Luo et al., 2007) and membrane to the Golgi. A recent study suggests that TMED10 RAB10 (Wang et al., 2010). cycles through the plasma membrane with improperly folded GPI- TMED10 has been observed to be localized at other intracellular anchored proteins (Zavodszky and Hegde, 2019). Hence, an compartments such as the ER, the ERGIC compartment, on interesting possibility would be that RAB21 is involved in secretory vesicles and at the plasma membrane (Pastor-Cantizano regulating the trafficking and degradation of improperly folded et al., 2016). Given the known functions of RAB21 in endocytosis GPI-anchored proteins, and as such RAB21 depletion would lead to and protein sorting (Jean et al., 2015; Pellinen et al., 2006; Del improper TMED10 localization. Olmo et al., 2019), and considering the observed RAB21-dependent In our recent proteomic study where we noticed RAB21/TMED10 TMED10 localization at the Golgi, it is also possible that RAB21 interaction, we did not observe any significant enrichment of could be involved in TMED10 recycling from the plasma TMED10 in APEX2:RAB21-mediated proximity labeling. This was Biology Open

4 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019 RESEARCH ARTICLE Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336 rather surprising given the large number of proximal RAB21 that this study provides a framework for further studies on the proteins identified in that study (Del Olmo et al., 2019). We believe association between the early endosomal RAB GTPase RAB21 and that this could be explained by the fact that APEX2 biotinylation p24 family members. occurs mostly on tyrosine residues (Lee et al., 2017), and p24 family members contain only short cytosolic domains with no tyrosine MATERIALS AND METHODS (Pastor-Cantizano et al., 2016). Therefore, although TMED10 could Cell culture still be proximal to APEX2:RAB21 in these experiments, it might All cell lines were grown in Dulbecco’s modified Eagle’s medium (DMEM) not be biotinylated and hence detected. This indicates the necessity supplemented with penicillin, streptomycin and 10% fetal bovine serum to combine experimental approaches to define protein interactomes. (Wisent) under 5% CO2 at 37°C. HeLa and HCT116 Flp-In/T-REx cells and In this regard, both SILAC and proximity labeling approaches will RAB21 knockout cell populations were described previously in (Del Olmo complement each other in future studies. et al., 2019). RAB21 modulates TMED10 and TMED2 protein levels. Generation of DNA constructs TMED10 was found to have a half-life of 3 h in the neurons (Liu PCDNA3-TMED10:3xHA was generated by amplifying TMED10 by PCR et al., 2008), while TMED2 was shown to have a very long half-life from HeLa cDNA generated using the Superscript III First-Strand synthesis in Vero cells (Füllekrug et al., 1999). TMED10 degradation in kit. The TMED10 PCR fragment was ligated into pCDNA3-3xHA using the neurons was mostly through the proteasome, while the mechanism In-Fusion cloning kit (Clontech). The GFP:TMED10 plasmid was a kind in Vero cells has not been characterized. Hence, it remains unclear gift of Robert Blum (Blum and Lepier, 2008). All constructs were validated whether a common pathway is responsible for the degradation of by sequencing. p24 family members. From our data, we could not identify the mechanism which contributes to the decreased abundance of Immunoprecipitations TMED10 and TMED2. We observed that TMED10 had a long half- 3.5×106 HCT116 or HeLa cells were plated in 100 mm dishes and grown for life in HeLa (>7 h) and we could not detect a shorter half-life in 24 h with 11 ng/ml of doxycycline to allow GFP:RAB21 induction. RAB21 knockout cells. Furthermore, proteasome or lysosome HCT116 cells were washed twice with cold PBS and lysed on ice for 20 min with 1 ml CoIP Buffer (1% IGEPAL CA-630, 1 mM EDTA, 150 mM NaCl, inhibition for 16 h did not strongly affect TMED10 levels. We did 0.1 mM EGTA, 25 mM Tris–HCl pH 7.4, 15 mM MgCl2, 2 mM Na3VO4, observe a slight increase in TMED10 levels upon lysosome 10% glycerol, 2× protease inhibitors) per plate. HeLa cells were fixed for inhibition (Fig. S3B), but this observation was not consistent over 15 min at room temperature with 0.5% formaldehyde with gentle rocking, the various repeats. Hence, decreased protein levels of TMED10 following which formaldehyde was quenched for 5 min with 125 mM and 2 could be attributed to an indirect effect on other p24 family glycine at room temperature. Fixed HeLa cells were washed twice with cold members or to a slight increase in degradation kinetics that could not PBS and lysed for 15 min at 4°C on a rotator with 1 ml of lysis buffer be observed in the timescale of our experiment. We ruled out a (1% IGEPAL CA-630, 1 mM EDTA, 150 mM NaCl, 25 mM Tris pH 7.4, general effect on global protein translation, since multiple 5% glycerol). For both HeLa and HCT116 cells, remaining membrane proteins are not downregulated in RAB21 knockout cells (Del aggregates and DNA were removed by centrifugation at 16,000 g for 12 min Olmo et al., 2019). at 4°C. Protein concentrations were determined using a BSA assay (Pierce), and immunoprecipitations were performed in individual tubes with TMED10 plays a dual role in cargo trafficking. By itself, equivalent quantities of proteins. 15 µl of GFP-Trap beads (ChromoTek) TMED10 is involved in specific secretion of GPI-anchored proteins were used for individual IPs. For HCT116 cells, immunoprecipitations were (Theiler et al., 2014) or PAR-2 (Zhao et al., 2014) towards the performed on a rotator for 2.5 h at 4°C. Beads were then washed twice with plasma membrane. On the other hand, TMED10 has been shown to CoIP buffer and twice again with CoIP buffer lacking IGEPAL CA-630. retain cargo such as MHC class I (Jun and Ahn, 2011) in the ER, and Immunoprecipitations in HeLa cells were carried out for 4 h on a rotator at PKC-δ in Golgi-like structures (Wang et al., 2011), thus inhibiting 4°C. Following this incubation, beads were washed three times with lysis their activities. Similarly, TMED10 is responsible for direct buffer. Finally, for both HCT116 and HeLa cells, excessive wash buffer was γ-secretase inhibition (Pardossi-Piquard et al., 2009). We removed from the beads at the end of the immunoprecipitation protocol and hypothesized that RAB21 could modulate general TMED10- 25 µl of 2× SDS loading buffer was added to each sample to elute proteins regulated cargo trafficking by disturbing TMED10 localization from beads. and stability. With respect to this hypothesis, RAB21 has been Pulldown assays shown to interact with Presenilin 1, reducing APP synthesis through 6 γ 3.5×10 HeLa cells were grown for 24 h in 100 mm plates, after which cells -secretase inhibition (Sun et al., 2017). Hence, it would be were lysed for 20 min on ice using 1 ml of MLB modified buffer (1% interesting to investigate if the involvement of RAB21 and IGEPAL CA-630, 10% glycerol, 100 µM EGTA and 100 µM GTP, 25 mM TMED10 in γ-secretase inhibition are linked or independent. HEPES, 150 mM NaCl, 20 mM MgCl2 and 1 mM sodium orthovanadate) Moreover, inhibition of TMED10 expression is involved in ATG4- supplemented with 2× protease inhibitors. Lysates were cleared by mediated autophagy activation (Shin et al., 2019). Hence, in centrifugation at 16,000 g for 12 min at 4°C. GST and GST:RAB21 were addition to its published role in modulating autophagosome- purified following this (Jean et al., 2012). Prior to the pulldowns, GST and lysosome fusion through VAMP8 trafficking (Jean et al., 2015), GST:RAB21 beads were washed three times in MLB modified buffer minus RAB21 could also influence other steps of autophagy via its IGEPAL CA-630 and incubated on a rotator at 4°C for 20 min in MLB interaction with TMED10. modified buffer lacking IGEPAL CA-630. Beads were further washed three In this study, we validated the interaction between RAB21 and times in complete MLB modified buffer and 900 µl of HeLa cell lysates was added to the beads for each pulldown, this was incubated for 1 h at 4°C on a TMED10, a cargo adaptor, and showed the role of RAB21 in rotator. Following this incubation, beads were washed three times with MLB modulating TMED10 localization. Considering that RAB21 and modified buffer containing 0.2% IGEPAL CA-630. Protein were eluted TMED10 are involved in the regulation of autophagy and with 30 µl of 2× SDS loading buffer. γ-secretase activity, respectively, and the fact that both of these pathways are associated with Alzheimer’s disease, it would be Immunoblots interesting to further investigate the functional relationship between For immunoblot analyses, 3×105 parental HeLa cells and 4.5×105 HeLa these two trafficking proteins in Alzheimer’s disease. We believe RAB21 KO cells were grown for 24 h in six-well plates. Cells were lysed Biology Open

5 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019 RESEARCH ARTICLE Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336 with 200 µl of CoIP buffer as described above for immunoprecipitations. relate and filter modules of Cell Profiler. Pearson correlations were also Lysates were quantified and the same amounts of proteins were used for generated using Cell Profiler, with the distinction that cells were manually analysis. Proteins were separated on 4–20% TGX precast gels (Bio-Rad) and identified. Immunoblot data densitometry and image processing were transferred onto PVDF membranes (Millipore) using the trans-blot turbo performed using Image Lab (Bio-Rad). Immunoblot images were cropped system from Bio-Rad. Antibodies used for immunoblotting were anti- and assembled using Photoshop and Illustrator respectively. For statistical RAB21 (1:1000, Invitrogen #PA5-34404), anti-GFP (1:500, Santa Cruz analyses of images or immunoblots, every sample was first subjected to a #9996), anti-TMED10 (1:1000, Abcam #134948), anti-TMED2 (1:1000, normality test (if a sufficient number of values were present), following Santa Cruz #376458), anti-GAPDH (1:8000, Cell Signaling #8884), anti- which either unpaired t-test, Mann–Whitney tests or one-sample t-tests LC3B (1:1000, Cell Signaling #3868), anti-TGN38 (1:1000, Santa Cruz were performed. #166594), anti-Ubiquitin (1:1000, Cell Signaling 3933), anti-Vps35 (1:500, Santa Cruz #374372) and anti-rabbit and mouse HRP (1:10,000, Jackson Quantitative PCR analysis Laboratories #115-035-144 and #115-035-146, respectively). Membranes Wild-type or RAB21 knockout cells were grown to 80% confluency in full were imaged on a Bio-Rad Chemidoc XR station following 5 min incubation media. CDNAs were prepared using the Maxima First Strand cDNA with Luminata Forte (Millipore) or Clarity Max chemiluminescent substrates synthesis kit for RT-qPCR with dsDNAse (Thermo Fisher Scientific) (Bio-Rad). On specific occasions, membranes were cut to allow probing with following the manufacturer’s instructions. Luna Universal qPCR Master multiple antibodies simultaneously. Mix was used for amplification and the reactions were performed on a Roche TMED10 stability experiments were performed by incubating cells in full LightCycler 96. Relative mRNA levels were calculated using the ΔΔCt media containing either 25 µg/ml cycloheximide or 10 µM MG132 for the method and normalized to GAPDH. TMED10, 2 and GAPDH primers were indicated amount of time. For Bafilomycin A1 treatments, 0.2 µg/ml and predesigned qPCR primers obtained from IDT. 0.1 µg/ml were used for the 4 h and the 16 h time point, respectively. A lower concentration was used for the 16 h time point due to Baf A1 toxicity. Acknowledgements Cells were lysed and proteins immunoblotted as described above. We are grateful to Robert Blum (Universitätsklinikum Würzburg) for kindly providing the TMED10:GFP (Blum and Lepier, 2008) construct used throughout this study. We

Immunofluorescence, colocalization and proximity thank Marie-Josee Boucher and Benoit Marchand for suggestions with the MG132 assay. We thank the proteomic and photonic microscopy cores at the Universite de ligation assay Sherbrooke for proteomic services and confocal use. Steve Jean and François- A total of 20,000 wild-type HeLa or 30,000 RAB21 knockout cells Michel Boisvert are members of the FRQS-Funded Centre de Recherche du CHUS. were plated on glass coverslips (#1.5) in 24-wells plate and cultured Steve Jean is a recipient of a Research Chair from the Centre de recherche me dicale overnight. The following day, pcDNA3-GFP:TMED10 or pcDNA3- de l’Universite de Sherbrooke (CRMUS). S.J. was supported by junior faculty salary TMED10:3xHA with or without pcDNA3-V5:RAB21 were transfected awards from CIHR and Fonds de Recherche du Que bec-Sante (FRQS). using Jetprime (Polyplus) following manufacturer’s instructions. 24 h following transfection, cells were washed twice with 1× PBS and fixed for Competing interests 15 min at room temperature with 250 µl of 4% paraformaldehyde in PBS. The authors declare no competing or financial interests. Cells were then washed three times 5 min each with 1× PBS. Fixed cells were blocked, and permeabilized for 60 min with 300 µl of 5% goat serum Author contributions and 0.3% Triton X-100 in PBS. Cells were then incubated overnight in a Conceptualization: T.D.O., S.J.; Methodology: T.D.O., C.L.-K., C.N., D.J., S.J.; humidified chamber at 4°C with primary antibody in 1× PBS containing Investigation: T.D.O., C.L.-K., C.N., S.J.; Resources: T.D.O., D.J., F.-M.B.; Data curation: T.D.O., C.L.-K., C.N.; Writing - original draft: T.D.O.; Writing - review & 0.3% Triton X-100 and 1% BSA. The following day, primary antibodies editing: T.D.O., S.J.; Supervision: S.J.; Project administration: S.J.; Funding were washed three times 5 min with 1× PBS. For immunofluorescence, cells acquisition: S.J. were incubated for 1 h in a humidified chamber with secondary antibody at room temperature in the same buffer as the primary antibody. Secondary Funding antibodies were washed three times for 5 min with 1× PBS at room This research was supported by operating grants from the Cancer Research Society temperature and cells were mounted in DAPI-containing mounting media (CRS; 22247) and the Canadian Institutes of Health Research (CIHR; 142305). (Sigma-Aldrich). For PLA, after having washed the primary antibodies, cells were incubated in a humidified chamber for 1 h at 37°C with Sigma probes (+) Supplementary information and (−) in the 1× Sigma dilution solution. Sigma probes were washed twice for Supplementary information available online at 5 min with 1× wash buffer A. The ligation step was performed at 37°C for http://bio.biologists.org/lookup/doi/10.1242/bio.045336.supplemental 30 min followed by two washes with buffer A. Amplification was performed for 100 min at 37°C in the dark. Cells were washed at room temperature twice References for 10 min each with 1× wash buffer B and finally for 1 min in 0.01× wash Alanko, J., Mai, A., Jacquemet, G., Schauer, K., Kaukonen, R., Saari, M., Goud, B. and Ivaska, J. (2015). Integrin endosomal signalling suppresses anoikis. Nat. buffer B and mounted in DAPI-containing mounting media (Sigma-Aldrich). Cell Biol. 17, 1412-1421. doi:10.1038/ncb3250 Antibodies used for immunofluorescences were anti-GM130 (1:100, Cell Anantharaman, V. and Aravind, L. (2002). The GOLD domain, a novel protein Signaling #12480), anti-ciMPR (1:100, Bio-Rad #MCA2048T) and anti- module involved in Golgi function and secretion. Genome Biol. 3, research0023. TGN46 (1:100, Novus Biological #NBP1-49643SS) and for PLA were anti- doi:10.1186/gb-2002-3-5-research0023 HA (1:1000, Cell Signaling #3724) and anti-V5 (1:5000, Sigma-Aldrich Barr, F. and Lambright, D. G. (2010). Rab GEFs and GAPs. Curr. Opin. Cell Biol. #V8012). 22, 461-470. doi:10.1016/j.ceb.2010.04.007 Blomen, V. A., Májek, P., Jae, L. T., Bigenzahn, J. W., Nieuwenhuis, J., Staring, J., Sacco, R., van Diemen, F. R., Olk, N., Stukalov, A. et al. (2015). Gene Image analysis and statistics essentiality and synthetic lethality in haploid human cells. Science 350, All images were acquired on an Olympus FV1000 using a 63× 1.42NA plan 1092-1096. doi:10.1126/science.aac7557 Apo N objective or on a Zeiss LSM880 using a 40× 1.4NA plan Apo Blum, R. and Lepier, A. (2008). The luminal domain of p23 (Tmp21) plays a critical objective. Imaging settings were selected to minimize pixel saturation and to role in p23 cell surface trafficking. Traffic 9, 1530-1550. doi:10.1111/j.1600-0854. ensure proper Pearson correlation calculation. For figure preparation, all 2008.00784.x microscopy images were tresholded and cropped on Adobe Photoshop and Burgo, A., Sotirakis, E., Simmler, M.-C., Verraes, A., Chamot, C., Simpson, assembled using Adobe Illustrator. All images were treated similarly, and J. C., Lanzetti, L., Proux-Gillardeaux, V. and Galli, T. (2009). Role of Varp, a only linear modifications were performed. The number of PLA puncta per Rab21 exchange factor and TI-VAMP/VAMP7 partner, in neurite growth. EMBO Rep. 10, 1117-1124. doi:10.1038/embor.2009.186 cell was established using Cell Profiler. Briefly, a pipeline allowing Burgo, A., Proux-Gillardeaux, V., Sotirakis, E., Bun, P., Casano, A., Verraes, A., transfected cell identification (using GFP signal from cotransfection of a Liem, R. K. H., Formstecher, E., Coppey-Moisan, M. and Galli, T. (2012). A small amount of pEGFP-C1 together with the other plasmids) was used and molecular network for the transport of the TI-VAMP/VAMP7 vesicles from cell the number of PLA puncta per EGFP-positive cell was established using the center to periphery. Dev. Cell 23, 166-180. doi:10.1016/j.devcel.2012.04.019 Biology Open

6 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019 RESEARCH ARTICLE Biology Open (2019) 8, bio045336. doi:10.1242/bio.045336

Contreras, I., Ortiz-Zapater, E. and Aniento, F. (2004). Sorting signals in the measurements in living cells using FRET. Dev. Cell 1, 139-153. doi:10.1016/ cytosolic tail of membrane proteins involved in the interaction with plant ARF1 and S1534-5807(01)00004-1 coatomer. Plant J. 38, 685-698. doi:10.1111/j.1365-313X.2004.02075.x Montesinos, J. C., Sturm, S., Langhans, M., Hillmer, S., Marcote, M. J., Contreras, F.-X., Ernst, A. M., Haberkant, P., Björkholm, P., Lindahl, E., Gönen, Robinson, D. G. and Aniento, F. (2012). Coupled transport of Arabidopsis p24 B., Tischer, C., Elofsson, A., von Heijne, G., Thiele, C. et al. (2012). Molecular proteins at the ER-Golgi interface. J. Exp. Bot. 63, 4243-4261. doi:10.1093/jxb/ recognition of a single sphingolipid species by a protein’s transmembrane domain. ers112 Nature 481, 525-529. doi:10.1038/nature10742 Pardossi-Piquard, R., Böhm, C., Chen, F., Kanemoto, S., Checler, F., Schmitt- Del Olmo, T., Lauzier, A., Normandin, C., Larcher, R., Lecours, M., Jean, D., Ulms, G., St George-Hyslop, P. and Fraser, P. E. (2009). TMP21 Lessard, L., Steinberg, F., Boisvert, F.-M. and Jean, S. (2019). APEX2- transmembrane domain regulates γ-secretase cleavage. J. Biol. Chem. 284, mediated RAB proximity labeling identifies a role for RAB21 in clathrin- 28634-28641. doi:10.1074/jbc.M109.059345 independent cargo sorting. EMBO Rep. 20, e47192. doi:10.15252/embr. Pastor-Cantizano, N., Montesinos, J. C., Bernat-Silvestre, C., Marcote, M. J. 201847192 and Aniento, F. (2016). p24 family proteins: key players in the regulation of D’Souza-Schorey, C. and Chavrier, P. (2006). ARF proteins: roles in membrane trafficking along the secretory pathway. Protoplasma 253, 967-985. doi:10.1007/ traffic and beyond. Nat. Rev. Mol. Cell Biol. 7, 347-358. doi:10.1038/nrm1910 s00709-015-0858-6 Füllekrug, J., Suganuma, T., Tang, B. L., Hong, W., Storrie, B. and Nilsson, T. Pellinen, T., Arjonen, A., Vuoriluoto, K., Kallio, K., Fransen, J. A. M. and Ivaska, (1999). Localization and recycling of gp27 (hp24gamma3): complex formation J. (2006). Small GTPase Rab21 regulates cell adhesion and controls endosomal with other p24 family members. Mol. Biol. Cell 10, 1939-1955. doi:10.1091/mbc. traffic of beta1-integrins. J. Cell Biol. 173, 767-780. doi:10.1083/jcb.200509019 ̈ 10.6.1939 Popoff, V., Adolf, F., Brugger, B. and Wieland, F. (2011). COPI budding within the Gommel, D., Orci, L., Emig, E. M., Hannah, M. J., Ravazzola, M., Nickel, W., Golgi stack. Cold Spring Harb. Perspect. Biol. 3, a005231. doi:10.1101/ Helms, J. B., Wieland, F. T. and Sohn, K. (1999). p24 and p23, the major cshperspect.a005231 transmembrane proteins of COPI-coated transport vesicles, form hetero- Rojas, A. M., Fuentes, G., Rausell, A. and Valencia, A. (2012). Evolution: the Ras oligomeric complexes and cycle between the organelles of the early secretory protein superfamily: evolutionary tree and role of conserved amino acids. J. Cell Biol. pathway. FEBS Lett. 447, 179-185. doi:10.1016/S0014-5793(99)00246-X 196, 189-201. doi:10.1083/jcb.201103008 Grosshans, B. L., Ortiz, D. and Novick, P. (2006). Rabs and their effectors: Shin, J. H., Park, S. J., Jo, D. S., Park, N. Y., Kim, J. B., Bae, J.-E., Jo, Y. K., Hwang, J. J., Lee, J.-A., Jo, D.-G. et al. (2019). Down-regulated TMED10 in achieving specificity in membrane traffic. Proc. Natl. Acad. Sci. USA 103, Alzheimer disease induces autophagy via ATG4B activation. Autophagy 15, 11821-11827. doi:10.1073/pnas.0601617103 1495-1505. doi:10.1080/15548627.2019.1586249 Hein, M. Y., Hubner, N. C., Poser, I., Cox, J., Nagaraj, N., Toyoda, Y., Gak, I. A., Simpson, J. C., Griffiths, G., Wessling-Resnick, M., Fransen, J. A. M., Bennett, Weisswange, I., Mansfeld, J., Buchholz, F. et al. (2015). A human interactome H. and Jones, A. T. (2004). A role for the small GTPase Rab21 in the early in three quantitative dimensions organized by stoichiometries and abundances. endocytic pathway. J. Cell Sci. 117, 6297-6311. doi:10.1242/jcs.01560 Cell 163, 712-723. doi:10.1016/j.cell.2015.09.053 Stenmark, H. (2009). Rab GTPases as coordinators of vesicle traffic. Nat. Rev. Mol. Hutagalung, A. H. and Novick, P. J. (2011). Role of Rab GTPases in membrane cell Biol. 10, 513-525. doi:10.1038/nrm2728 traffic and cell physiology. Physiol. Rev. 91, 119-149. doi:10.1152/physrev. Sun, Z., Xie, Y., Chen, Y., Yang, Q., Quan, Z., Dai, R. and Qing, H. (2017). Rab21, 00059.2009 a novel PS1 interactor, regulates γ-secretase activity via PS1 subcellular Jean, S. and Kiger, A. A. (2012). Coordination between RAB GTPase and distribution. Mol. Neurobiol. 55, 3841-3855. doi:10.1007/s12035-017-0606-3 phosphoinositide regulation and functions. Nat. Rev. Mol. Cell Biol. 13, 463-470. Theiler, R., Fujita, M., Nagae, M., Yamaguchi, Y., Maeda, Y. and Kinoshita, T. doi:10.1038/nrm3379 (2014). The α-helical region in p24γ2 subunit of p24 protein cargo receptor is Jean, S., Cox, S., Schmidt, E. J., Robinson, F. L. and Kiger, A. (2012). Sbf/ pivotal for the recognition and transport of glycosylphosphatidylinositol-anchored MTMR13 coordinates PI(3)P and Rab21 regulation in endocytic control of cellular proteins. J. Biol. Chem. 289, 16835-16843. doi:10.1074/jbc.M114.568311 remodeling. Mol. Biol. Cell 23, 2723-2740. doi:10.1091/mbc.e12-05-0375 Vicinanza, M., D’Angelo, G., Di Campli, A. and De Matteis, M. A. (2008). Function Jean, S., Cox, S., Nassari, S. and Kiger, A. A. (2015). Starvation-induced and dysfunction of the PI system in membrane trafficking. EMBO J. 27, MTMR13 and RAB21 activity regulates VAMP8 to promote autophagosome- 2457-2470. doi:10.1038/emboj.2008.169 lysosome fusion. EMBO Rep. 16, 297-311. doi:10.15252/embr.201439464 Wang, D., Lou, J., Ouyang, C., Chen, W., Liu, Y., Liu, X., Cao, X., Wang, J. and Jun, Y.-S. and Ahn, K.-S. (2011). Tmp21, a novel MHC-I interacting protein, Lu, L. (2010). Ras-related protein Rab10 facilitates TLR4 signaling by promoting β preferentially binds to 2-microglobulin-free MHC-I heavy chains. BMB Rep. 44, replenishment of TLR4 onto the plasma membrane. Proc. Natl. Acad. Sci. USA 369-374. doi:10.5483/BMBRep.2011.44.6.369 107, 13806-13811. doi:10.1073/pnas.1009428107 Lee, S.-Y., Kang, M.-G., Shin, S., Kwak, C., Kwon, T., Seo, J. K., Kim, J.-S. and Wang, H. B., Xiao, L. and Kazanietz, M. G. (2011). p23/Tmp21 associates with Rhee, H.-W. (2017). Architecture mapping of the inner mitochondrial membrane protein kinase Cδ (PKCδ) and modulates its apoptotic function. J. Biol. Chem. 286, proteome by chemical tools in live cells. J. Am. Chem. Soc. 139, 3651-3662. 15821-15831. doi:10.1074/jbc.M111.227991 doi:10.1021/jacs.6b10418 Yang, X., Zhang, Y., Li, S., Liu, C., Jin, Z., Wang, Y., Ren, F. and Chang, Z. (2012). Liu, S., Bromley-Brits, K., Xia, K., Mittelholtz, J., Wang, R. and Song, W. (2008). Rab21 attenuates EGF-mediated MAPK signaling through enhancing EGFR TMP21 degradation is mediated by the ubiquitin-proteasome pathway. internalization and degradation. Biochem. Biophys. Res. Commun. 421, 651-657. Eur. J. Neurosci. 28, 1980-1988. doi:10.1111/j.1460-9568.2008.06497.x doi:10.1016/j.bbrc.2012.04.049 Luo, W., Wang, Y. and Reiser, G. (2007). p24A, a type I transmembrane protein, Zavodszky, E. and Hegde, R. S. (2019). Misfolded GPI-anchored proteins are controls ARF1-dependent resensitization of protease-activated receptor-2 by escorted through the secretory pathway by ER-derived factors. eLife 8, e46740. influence on receptor trafficking. J. Biol. Chem. 282, 30246-30255. doi:10.1074/ doi:10.7554/eLife.46740 jbc.M703205200 Zhao, P., Metcalf, M. and Bunnett, N. W. (2014). Biased signaling of protease- Majoul, I., Straub, M., Hell, S. W., Duden, R. and Soeling, H.-D. (2001). KDEL- activated receptors. Front. Endocrinol. (Lausanne). 5, 67. doi:10.3389/fendo. cargo regulates interactions between proteins involved in COPI vesicle traffic: 2014.00067 Biology Open

7 Downloaded from http://bio.biologists.org/ at Universite de Sherbrooke on October 29, 2019