Lavoie Mathieu Phd 2015.Pdf

Université de Sherbrooke

Étude des mécanismes de dégradation sélective de l’ARN par la RNase III de Saccharomyces cerevisiae

Par Mathieu Lavoie Département de microbiologie et infectiologie

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 microbiologie

Thèse déposée le 7 janvier 2015 à Sherbrooke, Québec, Canada

Membres du jury d’évaluation Pr Brendan Bell, département de microbiologie et infectiologie Pr Sherif Abou Elela, département de microbiologie et infectiologie Pr Luc Gaudreau, département de biologie, Université de Sherbrooke Pr Marvin Wickens, Departement of Biochemistry, University of Wisconsin-Madison

Étude des mécanismes de dégradation sélective de l’ARN par la RNase III de Saccharomyces cerevisiae

Par Mathieu Lavoie Département de microbiologie et d’infectiologie

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 microbiologie, Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1H 5N4

RÉSUMÉ

Chez toutes les cellules, une modulation précise de l’expression des gènes est essentielle afin de réguler adéquatement leur métabolisme et de s’adapter aux changements environnementaux. En effet, c’est l’expression des gènes, plutôt que la séquence d’ADN, qui détermine en grande partie la diversité et la complexité des organismes. Celle-ci dépend principalement des changements dans les niveaux d’ARNs cellulaires résultant de la modification de l’équilibre entre leurs taux relatifs de synthèse et de dégradation. Alors que la régulation transcriptionnelle a été largement étudiée par le passé, des études récentes révèlent que la stabilité de l’ARN joue aussi un rôle important dans le modelage du transcriptome. Toutefois, les mécanismes qui assurent la dégradation précise et sélective des ARNs sont globalement mal compris.

Au cours de cette thèse, j’ai utilisé la ribonucléase III de levure Saccharomyces cerevisiae (Rnt1p) comme modèle pour étudier comment des transcrits spécifiques sont ciblés pour la dégradation et évaluer sa contribution à la régulation de l’expression génique. Les résultats indiquent que Rnt1p régule l’expression des gènes en utilisant une spécificité élargie pour des structures tige- boucles d’ARN. En effet, un nouveau motif structurel de Rnt1p permet la discrimination des tige-boucles ayant une séquence spécifique tout en bloquant la liaison à des hélices génériques d’ARN double-brin. D’un autre côté, l’identification des signaux de dégradation de Rnt1p à l’échelle du transcriptome a permis de révéler plus de 384 transcrits clivés par Rnt1p, dont la majorité sont des ARN messagers. En outre, l’impact de la délétion de RNT1 sur l’expression de ces gènes est influencé par les conditions de culture des cellules, ce qui suggère que Rnt1p est un important régulateur conditionnel de l’expression génique. Somme toute, les résultats présentés dans cette thèse démontrent comment des ARNs sont spécifiquement choisis pour la dégradation et soulignent l’importance de la dégradation nucléaire dans la régulation de l’expression génique en réponse à des changements environnementaux.

Mots clés = Rnt1p, RNase III, ARN double-brin, Saccharomyces cerevisiae, dégradation de l’ARN, régulation génique, réponse au glucose

Studies of the mechanisms of selective RNA degradation by the RNase III of Saccharomyces cerevisiae

By Mathieu Lavoie Département de microbiologie et d’infectiologie

Thesis presented to the Faculté de médecine et des sciences de la santé for the obtention of the title of philosophiae doctor (Ph.D.) in microbiology, Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1H 5N4

SUMMARY

Precise modulation of gene expression is essential for any cell in order to regulate its metabolism and adapt to environmental changes. In fact, it is gene expression, rather than DNA sequence alone, which mostly explains the functional diversity and complexity between the different cell types. As such, gene expression mainly results from changes in the levels of cellular RNAs which are, in turn, dependent on the equilibrium between their relative rates of synthesis and degradation. While transcriptional control has been largely studied in the past, recent publications reveal that changes in RNA stability also play an important role in shaping the transcriptome. Unfortunately though, the mechanisms ensuring precise and selective RNA degradation remains poorly understood.

In this thesis, I have used the yeast Saccharomyces cerevisiae ribonuclease III (Rnt1p) as a model to study how specific transcripts are targeted for degradation and evaluate its contribution to the regulation of gene expression. The results indicate that Rnt1p regulates gene expression using a broad specificity for structured RNA stem loops. Indeed, a new structural motif of Rnt1p permits discrimination of hairpins with specific sequence while blocking the binding of the generic RNA duplexes recognized by other members of the RNase III family. This highly specific mode of substrate recognition was found to be easily modulated by a flexible network of protein RNA interactions. On the other hand, transcriptome- wide identification of Rnt1p degradation signals uncovered more than 384 transcripts, including 291 mRNAs. Interestingly, the impact of RNT1 deletion on mRNA expression is modulated by changes in the growth conditions of the cell, indicating that Rnt1p is an important regulator of conditional gene expression. Overall, the results presented in this thesis demonstrate how specific RNAs are selected for degradation and highlight the importance of nuclear RNA decay for fine tuning gene expression in response to changes in growth conditions.

Key words = Rnt1p, RNase III, double-stranded RNA, Saccharomyces cerevisiae, RNA degradation, regulation of gene expression, glucose response

TABLE DES MATIÈRES

RÉSUMÉ ii

SUMMARY iii

TABLE DES MATIÈRES iv

LISTE DES TABLEAUX vii

LISTE DES FIGURES viii

LISTE DES ABRÉVIATIONS xi

INTRODUCTION 1

1. La dégradation de l’ARN joue un rôle important dans la régulation de l’expression génique chez la levure Saccharomyces cerevisiae 1 1.1. L’expression des gènes est un processus complexe et régulé à plusieurs niveaux 1 1.2. La dégradation de l’ARN : un processus bien conservé 4 1.3. Les principaux rôles de la dégradation de l’ARN chez S. cerevisiae 7 1.4. La régulation et la spécificité de la dégradation de l’ARN 10

2. Rnt1p, l’unique RNase III chez S. cerevisiae 12 2.1. Les RNase III forment une famille conservée et ubiquitaire d’endoribonucléases spécifiques pour l’ARN double-brin 12 2.2. La structure protéique de Rnt1p et des RNase III 13

3. Rnt1p influence plusieurs aspects de la biologie de l’ARN chez S. cerevisiae 16 3.1. Expression et maturation de l’ARN ribosomal 17 3.2. La maturation des ARNno et ARNnu 17 3.3. Le contrôle de la qualité des ARNs non-épissés 18 3.4. La dégradation conditionnelle des ARNm 18 3.5. Sauvegarde de la terminaison de la transcription de l’ARN polymérase II 20

4. Rnt1p reconnaît et clive une structure particulière de l’ARN double- brin 21 4.1. Trois épitopes sur l’ARN modulent la réactivité des substrats de Rnt1p 21 4.2. Le dsRDB de Rnt1p reconnaît la structure adoptée par la tétra- boucle 24

5. Objectifs du projet 26 v

CHAPITRE I : La ribonucléase III Rnt1p de Saccharomyces cerevisiae utilise un réseau de ponts hydrogènes pour lier et cliver ses substrats 28 Avant-propos 28 Résumé 29 Article 1 30

CHAPITRE II : Rnt1p utilise deux systèmes de mesure lors de la reconnaissance de ses substrats 70 Avant-propos 70 Résumé 71 Article 2 72

CHAPITRE III : Caractérisation des signaux de dégradation par la ribonucléase III Rnt1p à travers le génome de Saccharomyces cerevisiae 119 Avant-propos 119 Résumé 120 Article 3 121

CHAPITRE IV : Les produits de clivage d’un ARN messager rapporteur par Rnt1p sont exportés et dégradés au cytoplasme 177 Avant-propos 177 Résumé 178 Article 4 179

CHAPITRE V : Le couplage de la répression de la transcription et la dégradation de l’ARN permet la régulation de l’expression conditionnelle des gènes 206 Avant-propos 206 Résumé 207 Article 5 208

DISCUSSION 246

6. La reconnaissance des substrats par Rnt1p: entre flexibilité et spécificité 247 6.1. Rnt1p reconnaît la structure et la séquence de la tétra-boucle 247 6.2. La reconnaissance des substrats non-canoniques nécessite vraisemblablement un changement conformationnel de l’ARN 249 6.3. Rnt1p révèle un nouveau mode de reconnaissance de l’ARN par les RNase III 251 6.4. Un mécanisme d’action séquentiel assure la spécificité du clivage par Rnt1p 254 6.5. Différents éléments cis influencent la réactivité des substrats de Rnt1p 256

7. L’impact fonctionnel de Rnt1p sur l’expression des gènes 260 7.1. Rnt1p régule l’expression et clive plusieurs ARNm 260 7.2. La dégradation nucléaire est impliquée dans régulation conditionnelle de l’expression génique 262 7.3. Rnt1p cible les voies métaboliques sensibles aux conditions environnementales 265 7.4. Rnt1p influence la transcription de certains gènes 266 7.5. La liaison de l’ARN indépendante du clivage laisse entrevoir un rôle non-catalytique de Rnt1p 269

8. Les mécanismes de régulation de l’activité de Rnt1p 269 8.1. Les interactions protéines-protéines influencent le clivage de l’ARN par Rnt1p 269 8.2. Le recrutement co-transcriptionnel de Rnt1p 271 8.3. La localisation subcellulaire de Rnt1p permettrait la régulation conditionnelle des gènes en fonction du cycle cellulaire 273 8.4. Les conditions de culture affectent le niveau d’expression de Rnt1p 273 8.5. Rnt1p semble être régulé au niveau post-traductionnel 274

9. À propos de la classification des RNase III 275

CONCLUSION 278

REMERCIEMENTS 281

LISTE DES RÉFÉRENCES 282

ANNEXES 297

vii

LISTE DES TABLEAUX

INTRODUCTION Tableau 1 : Liste non-exhaustive des principaux facteurs impliqués dans la dégradation de l’ARN chez S. cerevisiae 6

CHAPITRE I Tableau 1 : Kinetic parameters of Rnt1p cleavage of bipartite substrates 38 Tableau 2 : Kinetic parameters of Rnt1p or ∆N-term cleavage of 2’-fluoro- modified substrates 46

CHAPITRE II Tableau 1 : Data collection and structure refinement statistics 79 Tableau S1 : Kinetic and binding parameters of Rnt1p proteins on the G2 substrates 105 Tableau S2 : List of oligonucleotides used for PCR-directed mutagenesis of Rnt1p 106

CHAPITRE III ** Les 19 tableaux supplémentaires qui accompagnent ce chapitre seront disponibles dans le fichier « Lavoie_Mathieu_PhD_2015_donnees.xls » soumis avec la thèse. **

CHAPITRE V Tableau S1 : List of yeast strains used in this study 243 Tableau S2 : List of oligonucleotides used in this study 244

ANNEXES Annexe 3 : Liste des gènes clivés par Rnt1p in vitro 300

viii

LISTE DES FIGURES

INTRODUCTION Figure 1 : Représentation simplifiée du dogme central de l’expression génique 3 Figure 2 : Voies générales et spécialisées de dégradation de l’ARN au cytoplasme 9 Figure 3 : La famille des RNase III 14 Figure 4 : Les différents rôles biologiques de Rnt1p 16 Figure 5 : Les RNAse III reconnaissent différentes structures d’ARNdb ayant peu de conservation de séquence 23 Figure 6 : Le dsRBD de Rnt1p reconnait la structure de la tétra-boucle NGNN 24

CHAPITRE I Figure 1 : Rnt1p cleavage of substrates with chemically modified stems 39 Figure 2 : Chemically modified guide RNAs enhance cleavage by Rnt1p in vivo 42 Figure 3 : Introduction of 2’-O-Me in Rnt1p primary binding site inhibits both binding and cleavage 44 Figure 4 : Rnt1p cleavage of RNA capped with 2’-F modified G2 tetraloop 47 Figure 5 : Rnt1p cleavage of RNA capped with 2’-F modified A1 tetraloop 52 Figure 6 : Hypothetical model for Rnt1p binding and cleavage 60 Figure S1 : Rnt1p cleaves enzymatically- and chemically-synthesized substrates with similar reactivity 68 Figure S2 : An example of the mobility shift assay (EMSA) used to calculate the dissociation coefficient of the different Rnt1p substrates 69

CHAPITRE II Figure 1 : Mode of substrate recognition by RNase III 76 Figure 2 : Cleavage site assembly of eukaryotic RNase III 80 Figure 3 : RNA-binding motif RBM0 identifies the NGNN tetraloop 82 Figure 4 : G clamp is tailored for the recognition of Gua16 86 Figure 5 : dsRBD interacts with the stem down to the ninth base pair below the tetraloop 89 Figure 6 : NTD dimer interacts with RNA to increase binding affinity and the precision of cleavage site selection 92 Figure 7 : Double-ruler mechanism for substrate selection by Rnt1p 98 Figure S1 : Composition and quality of RNA and protein molecules used in this study 107 Figure S2 : Sequence and structure of RNase III 109 Figure S3 : Structural details of Rnt1p-tetraloop interactions and impact of the Q373A mutation on substrate cleavage 110 Figure S4 : Recognition of the AGUC, AGAA, or AAGU tetraloop by ScRnt1p 111

Figure S5 : Distinct positions of the NTD dimer in relation to the RIIID dimer in KpDcr1 and ScRnt1p 112

CHAPITRE III Figure 1 : RNA degradation is induced by context dependent activation of randomly distributed cleavage motifs 128 Figure 2 : Gene expression is a poor indicator of substrate reactivity 131 Figure 3 : Gene expression independent identification of RNA degradation targets 134 Figure 4 : Sequence assisted identification of Rnt1p cleavage signals 138 Figure 5 : Evaluation of the different methods used for the detection of Rnt1p cleavage targets 141 Figure 6 : Rnt1p cleaves different classes of stem-loop structure with diverse sequence and structural requirements 144 Figure 7 : Deletion of RNT1 impairs the expression of in vitro substrates associated with respiration and carbohydrate metabolism 147 Figure S1 : In silico prediction of Rnt1p cleavage signals 165 Figure S2 : Detection and normalization of gene expression profiles 167 Figure S3 : Detection and validation of Rnt1p degradation targets 169 Figure S4 : Detection of Rnt1p cleavage sites using SALI 171 Figure S5 : Effect of changes in growth conditions on the expression of Rnt1p substrates 173 Figure S6 : Comparison of the advantages and disadvantages of each of the different methods used to determine Rnt1p cleavage targets 174

CHAPITRE IV Figure 1 : The fate of different Rnt1p cleaved RNA polymerase II transcripts 183 Figure 2 : Rnt1p cleaves reporter mRNAs in vivo 189 Figure 3 : Differences in the accumulation of cleaved reporter mRNAs in vivo correlate with differences in recognition by Rnt1p 191 Figure 4 : The unadenylated 5’ cleavage product is degraded by cytoplasmic RNases 194 Figure 5 : The uncapped 3’ cleavage product is degraded by the cytoplasmic Xrn1p 196

CHAPITRE V Figure 1 : Yeast RNase III selectively regulates the transcription factors associated with the Snf3p–Rgt2p glucose-signaling pathway 213 Figure 2 : Rnt1p optimizes the expression of the Rgt1-associated transcription factors 220 Figure 3 : Promoters partially mediate the Rnt1p-dependent repression of gene expression 222 Figure 4 : Rnt1p mediates cleavage site independent repression of gene expression 224 Figure 5 : Rnt1p enhances the glucose-dependent transcriptional repression of MTH1 227 Figure 6 : Proposed models of Rnt1p-dependent gene regulation 230 x

Figure S1 : Rnt1p cleavage is required for the conditional decay of MTH1 245

DISCUSSION Figure 7 : Rnt1p contacte la tétra-boucle et l’hélice d’ARNdb 252 Figure 8 : Mécanisme d’action de Rnt1p 256 Figure 9 : Modèles hypothétiques de régulation conditionnelle de l’expression génique par Rnt1p 264 Figure 10 : Sites potentiels de phosphorylation de Rnt1p 275

ANNEXES Annexe 1 : Résultats préliminaires de l’impact de la délétion de RNT1 sur l’ARN polymérase II 298 Annexe 2 : Résultats préliminaires de l’association de Rnt1p avec la chromatine 299 Annexe 4 : Essai de clivage par Rnt1p en présence d’extrait cellulaire total 301

LISTE DES ABRÉVIATIONS

ADN acide désoxyribonucléique ARN acide ribonucléique ARNdb ARN double-brin ARNi interférence par l’ARN ARN-ip immunoprécipitation de l’ARN ARNm ARN messager ARNnc ARN non-codant ARNno petit ARN nucléolaire ARNnu petit ARN nucléaire ARNr ARN ribosomal ARNt ARN de transfert BSB binding and stability box CEB cleavage efficiency box ChIP-seq immunoprécipitation de la chromatine couplé au séquençage à haut débit CTD domaine C-terminal de l’ARN polymerase II dsRBD domaine de liaison à l’ARN double-brin Endonucl endonuclease Exo exonuclease FISH fluorescence par hybridation in situ IBPB initial binding and positioning box MB middle box NMD non sense mediated decay NTD domaine N-terminal pol polymerase à ARN pré-ARNr transcript précurseur polycistronique de l’ARNr P-riche domaine riche en proline RBM motif de liaison à l’ARN RIIID domaine nuclease de la RNase III RISC RNA-induced silencing complex RNase ribonucléase RNase III ribonucléase III RNP ribonucléoprotéique RS-riche domaine riche en arginines et sérines SALI sequencing assisted loop identification siARN petit ARN interférant TRO Transcription run on UTR région transcrite non traduite d’un ARNm 3’ETS espaceur en 3’ de l’ARNr 2’-OH groupement 2’-hydroxyl

INTRODUCTION

1. La dégradation de l’ARN joue un rôle important dans la régulation de l’expression génique chez la levure Saccharomyces cerevisiae

1.1. L’expression des gènes est un processus complexe et régulé à plusieurs niveaux

En dépit des importantes différences morphologiques et fonctionnelles, toutes les cellules qui composent un individu partagent à la base le même contenu génétique, c’est-à-dire les mêmes molécules d’acide désoxyribonucléique (ADN). En fait, c’est l’expression différentielle des gènes, un processus par lequel l’information génétique encodée sur cet ADN est convertie en produits fonctionnels (protéines ou ARNs non-codants), qui permet à chaque cellule de se différencier, de remplir ses fonctions et de communiquer avec les autres cellules de l’organisme (Lodish et al., 2003). La régulation précise et adéquate de l’expression génique est donc essentielle au développement et à la survie des organismes.

L’expression des gènes comprend de nombreuses étapes (Figure 1). Elle débute avec la synthèse d’un ARN par la machinerie transcriptionnelle (polymérase à ARN). Cette étape est principalement influencée par le recrutement de divers facteurs de transcription au niveau des régions promotrices ou bien par des modifications épigénétiques de l’ADN, qui vont moduler de la fréquence d’initiation, la vitesse d’élongation et la terminaison de la transcription (Hughes et de Boer, 2013). Les ARNs fraîchement synthétisés subissent ensuite plusieurs modifications post-transcriptionnelles qui diffèrent selon le type de transcrit synthétisé (Alberts et al., 2002; Lodish et al., 2003). Par exemple, la majorité des ARNs pré-messagers (ARNm) contiennent des introns qui doivent être épissés pour former le transcrit mature. L’ajout d’une coiffe 7mGpppM en 5’ et la polyadénylation de l’extrémité 3’ des ARNm va également protéger ceux-ci de la dégradation par les exoribonucléases. Par ailleurs, notons que la dégradation de 2

l’ARN constitue un élément important de régulation post-transcriptionnelle en participant, entre autre, à la maturation des ARNs non-codants et à la destruction des transcrits aberrants ou indésirables (Parker, 2012). Par la suite, les ARNm correctement maturés et assemblés sous la forme de complexes ribonucléoprotéiques (RNP) sont exportés au cytoplasme où ils seront, ou bien traduits en protéines par les ribosomes, stockés pour une utilisation ultérieure, ou bien dégradés par les ribonucléases (Bond, 2006; Newbury, 2006; Shyu et al., 2008). Finalement, les protéines produites vont subir diverses modifications post- traductionnelles qui affecteront leur fonction et leur stabilité. Bref, l’expression des gènes est un processus complexe, hautement régulé et qui nécéssite la participation de nombreuses machineries et facteurs cellulaires.

Afin de répondre à ses besoins métaboliques et de s’adapter aux différents stress, la cellule va modifier l’expression de ses gènes notamment en modifiant son transcriptome, c’est-à-dire la nature et la quantité des ARNs cellulaires. Ultimement, la disponibilité de ces ARNs va dépendre de l’équilibre entre leur taux de production (transcription) et leur stabilité (dégradation) (Dolken et al., 2008; Elkon et al., 2010; Rabani et al., 2011) Alors que les mécanismes qui contrôlent la synthèse de l’ARN et la maturation de l’ARN ont été abondamment étudiés par le passé, la dégradation de l’ARN a longtemps été considérée comme un élément de régulation passif de l’expression globale des gènes et dont les rôles principaux étaient surtout d’assurer le recyclage des nucléotides, ainsi que la maturation et le contrôle de la qualité des ARNs (Garneau et al., 2007; Kuttykrishnan et al., 2010; Moore, 2002). Toutefois, des études récentes ont révélé qu’une grande partie des changements observés dans le transcriptome des cellules soumises à des conditions de stress sont dûes à des modifications de la stabilité de l’ARN plutôt qu’à l’activation de la transcription (Elkon et al., 2010; Munchel et al., 2011; Rabani et al., 2011; Romero-Santacreu et al., 2009; Shalem et al., 2008). En conclusion, la dégradation de l’ARN est un processus biologique important pour la régulation de l’expression des gènes qui contrôle à la fois la qualité et la quantité des ARNs cellulaires. 3

Figure 1 : Représentation simplifiée du dogme central de l’expression génique. Les gènes encodés sur l’ADN sont lus et transcrits en ARN par l’ARN polymérase II. Le pré-ARNm va ensuite subir plusieurs étapes de maturation (ajout de la coiffe, retrait des introns, polyadénylation, formation de complexes ribonucléoprotéiques) avant d’être exporté vers le cytoplasme. L’ARNm y sera traduit plusieurs fois en protéines par les ribosomes. Finalement, l’ARN est dégradé par des ribonucléases, permettant ainsi le recyclage des nucléotides.

1.2. La dégradation de l’ARN : un processus bien conservé

La dégradation de l’ARN est effectuée par des ribonucléases (RNases), qui sont des enzymes capables d’hydrolyser les liens phosphodiester qui unissent les ribonucléotides (Yang, 2011). Les RNases sont ubiquitaires à tous les domaines du vivant, de l’archaebactérie jusqu’à l’homme en passant par certains virus (Houseley et Tollervey, 2009; Read, 2013). Par ailleurs, il est fréquent de retrouver plusieurs RNases ayant des fonctions redondantes au sein d’un même organisme. Ainsi, la perte ou la mutation d’une seule ribonucléase n’abolit généralement pas complètement la dégradation de l’ARN (Houseley et Tollervey, 2009). Malgré cela, de nombreuses pathologies humaines, telles que la bêta-thalassémie (Frischmeyer et Dietz, 1999) ainsi que de nombreuses formes de cancer (Eick et al., 1985; Merritt et al., 2008; Michalova et al., 2013), ont été associées avec des défauts au niveau de la dégradation de l’ARN. Chez la levure Saccharomyces cerevisiae, la plupart des RNases ne sont pas essentielles à la survie, mais leur délétion peut engendrer des phénotypes spécifiques tels que la thermosensibilité, un ralentissement de la croissance ou une croissance filamenteuse (Catala et al., 2004; Kim, 2002). Le fait que les ribonucléases ne soient pas essentielles, mais que leur délétion mène à des phénotypes précis, suggère que la dégradation de l’ARN est importante pour la régulation de l’expression de gènes spécifiques et/ou dans des conditions particulières (Catala et al., 2004; Knight et Bass, 2001; Romero-Santacreu et al., 2009).

En plus de son omniprésence, la dégradation de l’ARN est un processus bien conservé à travers l’évolution. En effet, la plupart des ribonucléases et des cofacteurs identifiés chez la levure S. cerevisiae sont également présentes chez l’humain (Tableau 1) (Houseley et Tollervey, 2009). Les mécanismes de base ainsi que certains systèmes de régulation sont également bien conservés. Par exemple, le système de dégradation des ARNs non-sens (NMD; « nonsense mediated decay ») a été identifié entre autres chez les plantes, la mouche Drosophila melanogaster, les cellules de mammifères et les levures (Conti et Izaurralde, 2005; Shyu et al., 2008). Ce haut niveau de conservation fait en sorte que la levure S. 5

cerevisiae constitue un excellent modèle pour l’étude de la dégradation de l’ARN chez les cellules eucaryotes. De plus, la levure offre de nombreux avantages tels qu’un temps de division rapide, une grande disponibilité des souches, un génome entièrement séquencé, une littérature abondante et la facilité à réaliser des manipulations génétiques (Guthrie et Fink, 1991; Sherman, 2002). Par conséquent, la levure S. cerevisiae a été choisie comme modèle d’étude pour réaliser les travaux présentés dans cette thèse.

Tel que démontré dans le Tableau 1, la levure S. cerevisiae possède plusieurs ribonucléases qui peuvent être divisées en cinq catégories : les exonucléases 5’3’ qui dégradent l’ARN de manière processive à partir de l’extrémité 5’; les ribonucléases 3’5’ qui dégradent à partir de l’extrémité 3’; les endoribonucléases qui coupent à l’intérieur des transcrits; les déadénylases responsables de la dégradation des queues poly-A; et les enzymes de décoiffage qui retirent la coiffe en 5’ des ARN messagers (ARNm). Chez la levure, les principales RNases nucléaires sont Rrp6p, Rat1p et l’exosome (en association avec Rrp6p). En contrepartie, la majorité de l’ARN au cytoplasme est dégradée par Xrn1p et l’exosome (en association avec le complexe hélicase SKI). Ces RNases sont multifonctionnelles, c’est-à-dire qu’elles assument différents rôles en fonction des voies de dégradation dans lesquelles elles sont impliquées et de la nature du transcrit.

Tableau 1 : Liste non-exhaustive des principaux facteurs impliqués dans la

dégradation de l’ARN chez S. cerevisiae.

erme erme

) avec avec )

renf

SKI7

qualité et et qualité

, ,

A des ARNm. ARNm. des A ARNm des A

- -

unité unité

unités

osome nucléaire osome

ontrôle qualité ontrôle

de la de c

RRP44

NME1 MRP RNase à la

sous

spécifique à RNAse P à RNAse spécifique

et et

ontrôle qualité. Peut qualité. ontrôle

est la sous la est

SNM1

tés et chacune contient un un contient chacune et tés

RPR2

RPR1

et et

et linéarisation des lariats linéarisation et

sont les les sont

uni

) qui s’associe (via (via s’associe qui )

. Les autres protéines sont sont protéines autres . Les

NOT1

SKI2

RMP1

NOT1

échafaud

Principales fonctions / Description fonctions Principales transcription terminaison ARNnc, Maturation c ARNnc, Maturation l’ex réguler et avec s’associer Débranchement ARNnc Maturation des maturation la dans impliquée P RNase la dans impliquée MRP RNase ARNt. de la dégradation et 5.8S l’ARNr de maturation partagent MRP et P RNAse ARNm. quelques sous plusieurs ( ARN unité sous respectivement). que alors contrôle ARNnc, Maturation ARNs. des renouvellement catalytiques. activités les deux ( Hélicase son activité et régule cytoplasmique l’exosome ARNm coiffe des la de Retrait poly queues des Déadénylation CCR4 catalytiques. d’ accessoires poly queues des Déadénylation ARNs, des Renouvellement

» »

leader A A

- -

5’ 5’

boucles NG boucles

P protégé non OH protégé non OH P

- - - -

Cible 5’ 3’ Liens 2’ phosphodiester Tétra P = RNAse « séquence ARN des = MRP RNAse non signal exact A3 pré du Site défini. de et 5’UTR ARNr CLB2 3’ structuré ARN Coiffe 7mGpppN poly Queue poly Queue 5’

5’)

3’ 5’ 5’ 5’ 3’



    

Activité 5’ Exo 3’ Exo Endonucl. Endonucl. Endonuc 3’ Exo et Endonucl. Hélicase Décoiffage Dé adénylase 3’ (exo 3’ Exo 5’ Exo

Dégradation nucléaire Dégradation

Dégradation cytoplasmique Dégradation

Dégradation nucléaire et cytoplasmique et nucléaire Dégradation

3 9

- -

IV2L,WDR61

Homologues chez chez Homologues l’humain XRN2 EXOSC10 DBR1 trouvé non trouvé non trouvé non RRP29, RRP38, hPOP1, RPP21 RPP20, hPOP5, RPP21 trouvé non DIS3 EXOSC1 EXOSC4 SK DCP2 DCP1a/b, CNOT6L CNOT7 CNOT1 CNOT2 PAN2,3 XRN1

OT3,5, CDC36 OT3,5,

8, RPP1

43,45,46, MTR3 43,45,46,

Gènes chez chez Gènes S.cerevisiae RAT1 RRP6 DBR1 RNT1 RPR1 NME1 POP1,3 RPR2 SNM1 RMP1, RRP44 RRP40 RPP4, CLS4, RRP41 SKI7,2,3,8 DCP1,2 CCR4 POP2 NOT1 CAF16,40,120,130, N MOT2, PAN2,3 XRN1

NOT

Complexe P/MRP RNase Exosome SKI Complexe Décoiffage CCR4 PAN2/3 Les données sont tirées de (Houseley et Tollervey, 2009), de (Parker, 2012), ainsi que du site internet du NCBI (http://www.ncbi.nlm.nih.gov/homologene).

1.3. Les principaux rôles de la dégradation de l’ARN chez S. cerevisiae

Chez les organismes eucaryotes, la dégradation de l’ARN affecte tous les types d’ARNs cellulaires (Parker, 2012). En effet, de nombreux transcrits, tels que les ARNs ribosomaux (ARNr), les ARNs de transfert (ARNt), les petits ARN nucléaires (ARNnu) ou nucléolaires (ARNno), sont synthétisés sous la forme de précurseurs qui doivent être coupés et/ou partiellement dégradés pour générer le transcrit mature (Raijmakers et al., 2004). À noter que dans la plupart des cas, la maturation des ARNs non-codant à lieu dans le noyau des cellules. Bref, un rôle principal des ribonucléases est leur participation active à la maturation des différents transcripts cellulaires.

Un autre rôle de la dégradation de l’ARN est l’élimination des transcrits intergéniques, intragéniques ou antisens, fonctionnels ou non et qui résultent parfois de la transcription pervasive du génome (Wyers et al., 2005). Ces transcrits sont dégradés au noyau ou au cytoplasme par une variété de mécanismes, souvent redondants (Parker, 2012). Par exemple, certains CUTs (courts transcrits instables) qui sont normalement dégradés au noyau sous l’action de Rrp6p et du complexe TRAMP, sont aussi stabilisés en absence de la ribonucléase cytoplasmique Xrn1p (Thompson et Parker, 2007; van Dijk et al., 2011; Wyers et al., 2005).

En ce qui concerne les ARNm, on considère que leur principale voie de dégradation est cytoplasmique en raison de la présence de la coiffe 7mGpppM et de la queue poly-A qui les protègent contre la dégradation nucléaire (Garneau et al., 2007). En effet, pour la majorité des ARNm, la dégradation est initiée par la déadénylation des queues poly-A par les complexes CCR4-NOT ou PAN2/3 au cytoplasme (Figure 2A) (Garneau et al., 2007; Parker, 2012). Les ARNs sont ensuite dégradés directement par l’exosome à partir de l’extrémité 3’ ou bien par Xrn1p à l’extrémité 5’ suite au retrait de la coiffe par le complexe de décoiffage Dcp1p/Dcp2p. Dans certaines situations particulières, la dégradation 3’5’ ou le décoiffage peuvent aussi être effectués sans déadénylation préalable ou bien via 8

une coupure endonucléolytique à l’intérieur des transcrits (Badis et al., 2004). En conclusion, la stabilité des ARNm repose principalement sur la protection de leurs extrémités

En général, on considère que le rôle de la dégradation nucléaire dans la régulation de l’expression génique des ARNm se limite à la destruction des lariats suivant l’épissage des pre-ARNm et au contrôle de la qualité (Moore, 2002). En effet, il existe au noyau plusieurs mécanismes qui assurent la reconnaissance et la destruction des ARNm aberrants avant leur export vers le cytoplasme (Fasken et Corbett, 2009). En général, les pré-ARNm qui sont mals ou inefficacement maturés sont reconnus par des cofacteurs qui facilitent le recrutement des exoribonucléases. Alternativement, les pre-ARNm aberrants peuveut aussi être retenus au noyau lors de l’export afin de permettre la reconnaissance par la machinerie de dégradation (Fasken et Corbett, 2009). Naturellement, les transcrits qui présentent des anomalies grossières (absence de coiffe, épissage incomplet, etc.) seront directement dégradés par les exoribonucléases. Somme toute, la dégradation nucléaire constitue est un point de contrôle majeur qui assure la qualité des ARNm avant leur export vers le cytoplasme.

Le contrôle de la qualité des ARNm a également lieu au niveau du cytoplasme. En effet, des systèmes spécialisés, souvent associés avec la traduction, reconnaissent spécifiquement des ARN aberrants et engendrent leur dégradation (Figure 2B) (Schoenberg et Maquat, 2012). Le plus étudié est sans doute le NMD qui reconnaît les transcrits ayant un codon stop prémature, favorisant ainsi le recrutement du complexe de décoiffage et leur dégradation par Xrn1p. Originalement connu pour son rôle dans le contrôle des ARNs aberrants, le NMD peut aussi cibler certains transcrits « normaux » (Guan et al., 2006). Initiallement, il a été proposé que le NMD puisse être initié tant au cytoplasme que dans le noyau des cellules de mammifères (Culbertson et Neeno-Eckwall, 2005; Ishigaki et al., 2001). Des études subséquentes semblent toutefois indiquer que le NMD est restreint au cytoplasme (Kuperwasser et al., 2004; Trcek et al., 2013). 9

Figure 2 : Voies générales et spécialisées de dégradation de l’ARN au cytoplasme. (A) À la fin de leur vie utile, les ARNs sont déadénylés par les complexes CCR4-NOT ou bien PAN2/3. La coiffe 7mG est ensuite retirée par le complexe Dcp1p/Dcp2p (à gauche). L’extrémité 5’-phosphate ainsi générée est alors dégradée par l’exoribonucléase 5’3’ Xrn1p. La dégradation 3’5’ par l’exomome peut également avoir lieu sans décoiffage préalable (au centre). Les flèches en pointillés montrent des voies alternatives spécialisées indépendantes de la déadénylation. (B) Les principales voies cytoplasmiques de contrôle de la qualité des transcrits. Dans la dégradation non-sens (à gauche), la détection d’un codon stop prémature favorise, entre autre, le recrutement du complexe de décoiffage. Dans la voie de dégradation No-Go, le blocage des ribosomes en raison de la présence d’une structure stable de l’ARN induit une coupure endoribonucléolytique. En ce qui concerne la dégradation non- stop (à droite), l’absence de codon stop fait en sorte que les ribosomes se retrouvent au niveau de la queue poly-A, déplaçant ainsi les protéines qui lient cette dernière et facilitant la dégradation 3’5’. (Figure adaptée de Garneau et al., 2007 et de Parker R., 2012). 10

En conclusion, la dégradation nucléaire est principalement associée à la maturation des ARNs non-codants (ARNnc) et au contrôle de la qualité (Moore, 2002) alors que la dégradation cytoplasmique est surtout impliqué dans le renouvellement des transcrits et la régulation de l’expression génique. Étonnamment, la participation de la machinerie de dégradation nucléaire dans la régulation de l’expression des ARNm normaux demeure encore inexplorée.

1.4. La régulation et la spécificité de la dégradation de l’ARN

Tel que mentionné, la dégradation de l’ARN occupe de nombreuses fonctions et agit sur plusieurs types d’ARNs différents. De plus, chaque ARN dans une cellule est dégradé à un rythme (taux de demi-vie) qui lui est propre. Ce taux de demi-vie peut varier beaucoup d’un transcrit à l’autre, allant de quelques minutes jusqu’à plus d’une heure (Wang et al., 2002). Finalement, des modifications dans les conditions de croissance de la levure engendrent des changements systémiques, mais aussi spécifiques des taux de demi-vie de certains ARNm (Munchel et al., 2011; Wang et al., 2002). Toutes ces observations démontrent que la dégradation de l’ARN est un processus hautement régulé et spécifique. Malheureusement, les mécanismes qui controlent la précision, la spécificité et le « timing » de la dégradation de l’ARN sont loin d’être entièrement élucidés.

Un aspect important qu’il faut considérer pour bien comprendre la régulation de la stabilité des ARNs est que celle-ci est en constante compétition avec les autres processus cellulaires. Par exemple, la dégradation nucléaire est limitée par l’export rapide des ARNm (normaux) vers le cytoplasme (Libri et al., 2002; Zenklusen et al., 2002). Similairement, une modification du taux de traduction influence la stabilité cytoplasmique des transcrits (Hu et al., 2009). De manière générale, les ARNs incorrectement ou inefficacement transcrits, maturés, assemblés et exportés sont rapidement ciblés pour la dégradation (Schmid et Jensen, 2008). En conclusion, la régulation différentielle de chacun des processus impliqués dans l’expression génique, en réponse à un stress par exemple, est 11

susceptible d’influencer également la stabilité des ARNs de manière globale et/ou spécifique.

Une deuxième observation importante est que la majorité de l’ARN chez S. cerevisiae est dégradé par des exonucléases peu spécifiques. Par conséquent, la protection des extrémités 5’ et 3’ constitue un point de contrôle déterminant dans la stabilité globale des transcrits. Au cours des dernières années, de nombreuses études ont révélé une panoplie de systèmes de régulation et de cofacteurs qui influencent la stabilité de l’ARN, notamment en stimulant ou réprimant le décoiffage et la déadénylation des ARNm. Une revue exhaustive de ces mécanismes de régulation a été présenté par Roy Parker en 2012 (Parker, 2012). Globalement, il faut retenir que ces facteurs peuvent agir à plusieurs niveaux, soit en rendant accessibles les extrémités normalement protégées par des protéines, en favorisant le recrutement des déadénylases ou des complexes de décoiffage, ou bien en stimulant directement l’activité de ces derniers ou des exoribonucléases. Certains cofacteurs sont généraux, comme par exemple la protéine Pab1p qui prévient la déadénylation des queues poly-A des ARNm (Tucker et al., 2002), alors que d’autres sont spécifiques à certains transcrits, comme les protéines Puf qui lient une séquence particulière dans le 3’ UTR d’un ensemble restreint d’ARNm, stimulant ainsi leur dégradation (Goldstrohm et al., 2007). En conclusion, la stabilité des ARNs est largement contrôlée par la composition des complexes ribonucloprotéiques (RNP) dont ils font partie ainsi que le recrutement de cofacteurs qui influencent l’activité des ribonucléases. Malheureusement, les signaux et les mécanismes qui permettent l’activation conditionnelle de ces cofacteurs sont mal compris.

Finalement, une autre façon d’assurer la dégradation conditionnelle et spécifique d’un transcrit est la voie endonucléolytique (Figure 2A). Chez les eucaryotes supérieurs, il existe de nombreuses endoribonucléases qui sont capables d’initier la dégradation de transcrits distincts en coupant à l’intérieur de ceux-ci (Li et al., 2010a). Le mécanisme le plus connu est sans doute l’interférence par l’ARN (ARNi). Dans ce mécanisme, la coupure endonucléolytique des ARNm 12

ciblés est effectuée par la protéine AGO2 du complexe RISC. L’ARNi requiert également la participation des endoribonucléases III (RNase III), Dicer et Drosha (Ameres et al., 2007). Le mécanisme de RNAi est absent chez S. cerevisiae (Chang et al., 2012). Toutefois, il semblerait que Rnt1p, l’orthologue de Drosha, est également capable de couper spécifiquement l’ARNm de certains gènes de la levure (Ge et al., 2005; Lee et al., 2005). Il n’est toutefois pas clair si la régulation de l’expression des ARNm par Rnt1p est significative, comme dans l’exemple du RNAi, ou marginale.

2. Rnt1p, l’unique RNase III chez S. cerevisiae

2.1. Les RNase III forment une famille conservée et ubiquitaire d’endoribonucléases spécifiques pour l’ARN double-brin

La famille des RNase III comprend les endoribonucléases spécifiques à l’ARN double-brin (ARNdb) impliquées dans de nombreux aspects de la biologie de l’ARN telle que la maturation et la régulation de l’expression de plusieurs ARNs codants et non-codants. La première RNase III a été purifiée et caractérisée à partir de la bactérie Escherichia coli par le groupe de Norton Zinder en 1968 (Robertson et al., 1968), soit quelques années seulement après la publication de l’emblématique mécanisme d’action de la RNase A par Tony Mathias et Bob Rabin (Findlay et al., 1962). Déjà à cette époque, les auteurs avaient remarqué la haute spécificité des RNase III pour l’ARNdb. Aujourd’hui, nous savons que les RNase III sont ubiquitaires chez les bactéries et les organismes eucaryotes (Court et al., 2013). Des analyses par alignement de séquence ont aussi révélé la présence potentielle de RNase III dans le génome de certaines archaebactéries (Li et al., 2010b), mais la fonction et l’activité chez ces organismes n’a pas encore été démontrée. Chez les eucaryotes, la première RNase III identifiée fut l’enzyme Pac1 chez la levure Schizosaccharomyces pombe en 1991 (Iino et al., 1991). De son côté, le génome de la levure S. cerevisiae encode un seul réprésentant de la famille des RNase III, appelé Rnt1p. Rnt1p est une protéine de 471 acides aminés 13

ayant un poids moléculaire de 54,1 kilodalton qui a été identifiée pour la première fois en 1996 dans le laboratoire de Manuel Ares Jr. où les auteurs ont démontré son rôle dans la maturation de l’ARNr (Abou Elela et al., 1996).

2.2. La structure protéique de Rnt1p et des RNase III

Les RNase III présentent une grande diversité de structures protéiques. Par conséquent, la famille est divisée en quatre classes selon les domaines retrouvés chez ses membres (Figure 3) (Lamontagne et al., 2001). La classe I regroupe les plus petites RNase III qui sont généralement retrouvées chez les bactéries. Ces dernières participent à la maturation des ARNr et la régulation de l’expression génique de certains ARNm (Court et al., 2013). La classe II regroupe les enzymes Rnt1p et Pac1p retrouvées chez les levures S. cerevisiae et Schizosaccharomyces pombe, respectivement. Rnt1p est impliqué notamment dans la maturation de divers ARNs non-codants et la dégradation de quelques ARNm et pré-messagers (Abou Elela et al., 1996; Danin-Kreiselman et al., 2003; Ge et al., 2005; Ghazal et al., 2005). La structure de Rnt1p se distingue des RNAse III bactériennes par la présence d’une extension N-terminale (NTD) apparemment dépourvue de domaines protéiques classiques ainsi que d’une courte extension C-terminale (Lamontagne et Abou Elela, 2001). Le NTD participe à la dimérisation intra- et intermoléculaire de Rnt1p, contribue à stabiliser le complexe enzyme:substrat et contiendrait un signal de localisation nucléaire (Lamontagne et al., 2000). Un autre signal de localisation nucléaire est situé au niveau de l’extension C-terminale de la protéine (Catala et al., 2004). En conséquence, Rnt1p présente une localisation nucléaire. De manière plus précise, Rnt1p s’accumule principalement au niveau du nucléole, ce qui corrèle bien avec son rôle dans la maturation de l’ARNr (Abou Elela et al., 1996). De plus, l’accumulation au nucléole ou au nucléoplasme semble régulée en fonction du cycle cellulaire (Catala et al., 2004). Finalement, les enzymes des classes III et IV, représentées respectivement par Dicer et Drosha sont retrouvées chez les eucaryotes pluricellulaires comme l’humain et montrent une duplication de leur RIIID. Elles s’accompagnent aussi de divers domaines tels que le domaine PAZ important pour la reconnaissance des substrats de Dicer 14

(MacRae et al., 2006) ou des domaines P-riche ou RS-riche vraisemblablement impliqués dans les interactions protéine-protéine chez Drosha (Lau et al., 2012; Nicholson, 2014). Le rôle principal de Drosha et de Dicer est de participer à la maturation des petits ARNs régulateurs (microARNs et siARN) (Johanson et al., 2013; Yeom et al., 2006).

Figure 3 : La famille des RNase III. Schéma illustrant les principaux domaines protéiques retrouvés chez les archétypes des différentes classes de RNase III. Certaines variations de structures peuvent toutefois survenir à l’intérieur d’une même classe. Par exemple, l’enzyme Dicer (classe IV) de Giardia internalis ne possède pas de domaine hélicase ni de dsRBD (MacRae et al., 2006). NLS= signal de localisation nucléaire; dsRBD= domaine de liaison à l’ARN double-brin; RIIID= domaine nucléaire; NTD= domaine N-terminal; RS-riche= région riche en arginines et sérines; P-riche= région riche en prolines. Les diagrammes ne sont pas à l’échelle.

Tel que présenté à la Figure 3, toutes les RNase III possèdent au moins un domaine nucléase caractéristique, appelé le « RNase III domain » ou RIIID, qui est responsable de l’hydrolyse des liens phosphodiesters de l’ARN au site de coupure. Le domaine RIIID contient notamment la séquence signature (ERLEFLGD) propre aux RNase III (Filippov et al., 2000; Gan et al., 2006). Le RIIID fonctionne sous la forme de dimère relié par une interface hydrophobique (Blaszczyk et al., 2001). Conséquemment, Rnt1p et les RNase III bactériennes sont généralement retrouvées sous la forme d’homodimères, chaque sous-unité comportant un site catalytique. L’alignement des dimères de RIIID fait en sorte que chacun des brins 15

de l’hélice d’ARNdb est coupé tout en générant des produits ayant une extension de deux nucléotides à l’extrémité 3’ (Blaszczyk et al., 2001). Le mécanisme catalytique de Rnt1p, tout comme celui des autres membres de la famille RNase III, est similaire à celui de la RNase H et nécessite la participation de deux métaux divalents (Mg2+, principalement). Ces cofacteurs sont essentiels pour l’activation du nucléophile (eau) responsable de l’attaque du lien P-O 3’, ce qui génère des produits 5’-phosphate et 3’-hydroxyl (Gan et al., 2006; Yang, 2011). La coordination des métaux divalent et la formation du centre catalytique est assuré par la présence de quatre acides aminés hautement conservés au niveau du RIIID, dont deux sont présents dans la séquence signature (soulignés ci-haut). Au cours de ce projet de recherche, nous découvrirons que deux résidus supplémentaires participent aussi à la catalyse chez Rnt1p (voir au chapitre II). Chez la bactérie, le RIIID en soi est capable de cliver l’ARN de manière spécifique, ce qui indique que ce domaine est capable de lier l’ARNdb (Sun et al., 2001; Weinberg et al., 2011). À l’opposé, la liaison stable des substrats par Rnt1p requiert la participation d’un autre domaine protéique, le domaine de liaison à l’ARNdb (dsBRD) (Lamontagne et al., 2000).

Rnt1p, tout comme la grande majorité des RNase III, possède un dsRBD à son extrémité C-terminale qui est relié au RIIID par un lien flexible. Le dsRBD est un motif bien connu et conservé qui se retrouve dans de nombreuses protéines liant l’ARN telles que les protéines ADARs et Staufen (Doyle et Jantsch, 2002). Comme son nom l’indique, la principale fonction de ce domaine est de s’associer de manière spécifique à l’ARN double-brin en reconnaissant à la fois la structure générale de l’hélice de type A et en interagissant de manière intensive avec les groupements 2’-hydroxyl du sillon mineur de l’ARN (Chang et Ramos, 2005). Conséquemment, le dsRBD a très peu d’affinité pour l’ADN ou pour l’ARN simple- brin. Nous verrons à la section 4.2 comment ce domaine influence la reconnaissance des substrats par Rnt1p.

3. Rnt1p influence plusieurs aspects de la biologie de l’ARN chez S. cerevisiae

Rnt1p n’est pas une protéine essentielle à la survie de S. cerevisiae. Toutefois, la délétion du gène RNT1 (YMR239C) ou l’inactivation de la protéine engendre de multiples défauts phénotypiques comme, par exemple, une croissance thermosensible et ralentie, un délai du cycle cellulaire et de la division nucléaire, des malformations morphologiques ou encore un allongement des séquences télomèriques (Abou Elela et Ares, 1998; Catala et al., 2004; Larose et al., 2007). Ces effets pléiotropiques témoignent des rôles importants et variés qu’occupe Rnt1p au sein de la levure (Figure 4).

Figure 4 : Les différents rôles biologiques de Rnt1p. (A) La coupure par Rnt1p d’une structure tige-boucle dans la région 3’ETS du pré-ARNr 35S constitue la première étape de la maturation de l’ARNr 25S. (B) Rnt1p clive l’ARN en 5’ des ARNno (parfois en 3’) et en 3’ des ARNnu. (C) Rnt1p coupe dans la région intronique de certains transcrits, tels que RPL18a, lorsque ceux-ci sont incorrectement épissés. E1 et E2 désignent les exons. (D) Une structure tige-boucle à l’intérieur de la séquence codante de certains ARNm matures, tels que celui du gène MIG2, est reconnue et clivée par Rnt1p pour initier la dégradation du transcrit. (E) Rnt1p est impliqué dans la sauvegarde de la terminaison de la transcription. Rnt1p coupe le transcrit polycistronique généré lorsque l'ARN polymérase II (pol) ne s’arrête pas au site de polyadénylation (pA) attendu. La coupure va ensuite provoquer le décrochement de l’ARN polymérase. Des détails supplémentaires sont fournis dans le texte principal. 17

3.1. Expression et maturation de l’ARN ribosomal

À l’instar des RNase III bactériennes, Rnt1p joue un rôle important dans la maturation du précurseur de l’ARN ribosomal (pré-ARNr). Chez les eucaryotes, 3 des 4 ARNr sont transcrits par l’ARN polymérase I sous la forme d’un long précurseur qui est ensuite coupé et maturé pour produire les ARNr 25S, 18S et 5.8S matures. Or, il a été démontré que Rnt1p reconnaît et clive une structure tige- boucle surmontée d’une tétra-boucle AGGA située au niveau de l’extrémité 3’ du pré-ARNr (3’ETS) (Figure 4A) (Abou Elela et al., 1996). Ainsi, l’inactivation d’un allèle thermosensible de RNT1 cause une diminution des ARNr matures et l’accumulation du pré-ARNr 35S. La délétion du gène RNT1 diminue également la synthèse de nouveaux ARNr en plus de modifier la structure de la chromatine au locus ribosomal (Abou Elela et al., 1996; Catala et al., 2008). De plus, Rnt1p interagit physiquement avec deux sous-unités du complexe de l’ARN polymérase I et cette interaction est importante afin d’assurer la maturation adéquate du pré- ARNr (Catala et al., 2008). Finalement, il est proposé que le clivage du 3’ETS par Rnt1p est important pour assurer la terminaison adéquate de la transcription de l’ARN polymérase I (Braglia et al., 2011). Pour conclure, Rnt1p a une influence majeure à la fois sur la synthèse et la maturation des ARNr et plusieurs indices suggèrent que le clivage se produit de manière co-transcriptionnelle.

3.2. La maturation des ARNno et ARNnu

Les petits ARNs nucléolaires (ARNno) ont pour rôle de guider la modification chimique (méthylation et pseudouridylation) de bases spécifiques sur d’autres ARNs, notamment l’ARNr. Ils sont transcrits par l’ARN polymérase II sous la forme de transcrits indépendants, polycistroniques ou alors encodés dans certains introns (Chanfreau et al., 1998b; Lee et al., 2003; Qu et al., 1999). De manière analogue à l’ARNr, Rnt1p reconnaît et coupe une structure tige-boucle de type NGNN située en amont (généralement) ou en aval d’une majorité ARNno matures (Figure 4B) (Chanfreau et al., 1998a). Cette étape est souvent essentielle à la formation des ARNnu et ARNno matures. Elle permet aussi la séparation en 18

unités individuelles des ARNno polycistroniques ainsi que la maturation des ARNno présents dans les introns (Qu et al., 1999). Une recherche génomique pour les signaux de clivage a révélé que Rnt1p participe à la maturation d’au moins 35 ARNno (Chanfreau et al., 1998a; Chanfreau et al., 1998b; Davis et Ares, 2006; Ghazal et al., 2005; Kufel et al., 2000; Lee et al., 2003; Qu et al., 1999).

Rnt1p est aussi impliqué dans la maturation des petits ARNs nucléaires (ARNnu), qui sont des composantes importantes du splicéosome (Figure 4B). Le génome de S. cerevisiae encode cinq ARNnu qui sont transcrits par l’ARN polymérase II. Quatre d’entre eux (U1, U2, U4 et U5) présentent une structure tige- boucle de type NGNN à l’extrémité 3’ de leur transcrit et qui est clivée par Rnt1p (Abou Elela et Ares, 1998; Allmang et Tollervey, 1998; Chanfreau et al., 1997; Seipelt et al., 1999). En conclusion, un des rôles cellulaire principal de Rnt1p est la maturation des ARNnc.

3.3. Le contrôle de la qualité des ARNs non-épissés

De nombreux systèmes de contrôle assurent la dégradation des pré-ARNm qui n’ont pas été correctement épissé (Parker, 2012). Étonnamment, il a été observé que la délétion du gène RNT1 cause l’accumulation d’un certain nombre de pré-ARNm non-épissés. De plus, Rnt1p clive in vitro des tétra-boucles ayant une séquence GGUU qui correspondent à celles retrouvées dans les régions introniques des gènes RPS22B et RPL18A (Figure 4C). Le clivage par Rnt1p va ensuite initier la dégradation complète des transcrits aberrants par les voies exonucléolytiques (Danin-Kreiselman et al., 2003). Un mécanisme similaire a été observé pour le gène MATa1 (Egecioglu et al., 2012). Ces résultats suggèrent que Rnt1p joue un rôle dans le contrôle nucléaire de la qualité des ARNm et la dégradation des lariats.

3.4. La dégradation conditionnelle des ARNm

L’implication des RNase III dans la régulation de l’expression des ARNm est particulièrement bien caractérisée chez E. coli et les eucaryotes supérieurs 19

(interférence par l’ARN, par exemple) (Court et al., 2013; Filipowicz et al., 2008). Ce rôle est toutefois moins bien étudié dans le cas de Rnt1p. Des analyses effectuées avec des puces à ADN ont révélé que l’expression de plus de 450 gènes est augmentée plus de deux fois en absence du gène RNT1, suggérant que Rnt1p régule la stabilité de ces transcrits (Ge et al., 2005; Lee et al., 2005). Toutefois, il est vite apparu que plusieurs ARNm surexprimés in vivo ne sont pas nécessairement coupés par Rnt1p in vitro, ce qui rend difficile la distinction entre l’impact direct ou indirect de Rnt1p sur le niveau d’expression de ces gènes.

Outre les travaux présentés dans cette thèse, le clivage direct par Rnt1p des ARNm matures a seulement été démontré pour huit gènes jusqu’à maintenant : SWI4 et HSL1 (impliqués dans le contrôle du cycle cellulaire en réponse au stress), EST1 (sous-unité de la télomérase), ADI1 (voie de sauvetage de la méthionine), FIT2 et ARN2 (voie de transport du fer), BDF2 (remodelage de la chromatine et transcription), et MIG2 (voie de signalisation du glucose) (Catala et al., 2012; Ge et al., 2005; Larose et al., 2007; Lee et al., 2005; Roy et Chanfreau, 2014; Zer et Chanfreau, 2005). On présume que la coupure de Rnt1p à l’intérieur de la séquence codante (FIT2 serait coupé au niveau du 5’UTR) provoque la dégradation complète des transcrits par les exoribonucléases (Figure 4D). Il est intéressant de constater que tous ces gènes codent pour des protéines qui peuvent se retrouver au noyau, ce qui suggère que la sélection des substrats est influencée par la localisation cellulaire des produits qu’ils encodent. On notera également que la plupart de ces gènes sont exprimés de manière conditionnelle dans les cellules (cycle cellulaire, réponse au glucose, etc) et que Rnt1p réprime leur expression dans des conditions normales de croissance. De plus, il a été démontré que la délétion de RNT1 affecte la réponse cellulaire face à différents stimuli et, dans le cas de MIG2, retarde l’élimination conditionnelle de l’ARNm (Ge et al., 2005). Ces résultats suggèrent donc que Rnt1p est un important régulateur de l’expression de gènes exprimés conditionnellement.

3.5. Sauvegarde de la terminaison de la transcription de l’ARN polymérase II

Plus récemment, un nouveau rôle a été attribué à Rnt1p, soit la sauvegarde de la terminaison de la transcription de l’ARN polymérase II indépendante de la polyadénylation. Pour quelques gènes, il a été observé que la délétion de RNT1 cause une accumulation de transcrits polycistroniques (les gènes NPL3-GPI17, par exemple) ou ayant une extrémité 3’ allongée (RPL8a ou NAB2) (Ghazal et al., 2009; Lee et al., 2005; Rondon et al., 2009). Des essais d’immunoprécipitation de la chromatine et d’élongation de la transcription TRO (« transcription run on ») ont montré qu’en absence de Rnt1p, la transcription par l’ARN polymérase II se poursuit au-delà du site de polyadénylation habituel (Rondon et al., 2009). Le transcrit alternatif qui en résulte contient alors une tétra-boucle de type NGNN normalement encodée dans l’espace intergénique. La coupure de cette tétra- boucle par Rnt1p génère un point d’accès pour Rat1p qui peut alors dégrader le bout d’ARN 5’ en cours de transcription et ainsi provoquer la terminaison de la transcription (Figure 4E). Ce mécanisme serait analogue au mécanisme normal de terminaison appelé le modèle torpille. Rnt1p semble donc constituer un système de sauvegarde de la terminaison de la transcription lorsque celle-ci n’est pas effectuée au site poly-A canonique (Ghazal et al., 2009; Rondon et al., 2009).

En résumé, Rnt1p semble impliqué dans plusieurs mécanismes très différents tels que la maturation des ARNnc, la régulation de l’expression des ARNm, le contrôle de la qualité des pré-ARNm et la terminaison de la transcription. De plus, on remarque que Rnt1p semble réguler des transcrits ayant des structures et des fonctions très différentes. En effet, Rnt1p reconnaît des courts ARNs non-codants tout comme des longs ARNm et même des transcrits polycistroniques. La coupure peut avoir lieu dans les régions intergéniques, ou bien les régions UTR ou bien à l’intérieur même des régions codantes. En conséquence, le rôle et la fonction principale de Rnt1p dans la régulation de l’expression ainsi que l’ensemble des règles qui dictent la sélection des substrats ne sont pas bien définis. Jusqu’ici, la seule caractéristique commune entre tous les substrats connus de Rnt1p est qu’ils présentent une structure tige-boucle de type « NGNN ». 21

4. Rnt1p reconnaît et clive une structure particulière de l’ARN double-brin

À ce jour, la presque totalité des substrats connus de Rnt1p présentent une structure tige-boucle surmontée de quatre nucléotides (tétra-boucle) et dont la deuxième position est toujours occupée par une guanine (Figure 5A). Les deux seules exceptions à cette règle sont snR38 qui montre une tétra-boucle AAGU (Ghazal et Abou Elela, 2006) et MATa1 qui présente une boucle à sept nucléotides (Egecioglu et al., 2012). Les substrats de Rnt1p montrent aussi une certaine conservation pour une adénosine à la première position des tétra-boucles NGNN, bien qu’elle ne soit pas essentielle (Seipelt et al., 1999). Les positions trois et quatre ne montrent aucune conservation particulière de séquence. Finalement, le clivage a généralement lieu à 14 et 16 nucléotides à partir de la tétra-boucle sur chaque brin respectif de l’ARNdb (Lamontagne et al., 2001). Le mécanisme qui dicte la position des sites de clivage par rapport à la boucle est encore inconnu.

4.1 Trois épitopes sur l’ARN modulent la réactivité des substrats de Rnt1p

Rnt1p est la seule RNase III à reconnaître spécifiquement des tétra-boucles NGNN. En effet, les substrats des autres RNase III montrent généralement peu ou pas de conservation de séquence et la structure globale varie d’une enzyme à l’autre (Figure 5B). Des études biochimiques utilisant des tétra-boucles de longueur, séquence et structure variées ont permis d’identifier trois régions (épitopes) de la tige-boucle qui contrôlent la liaison de Rnt1p et/ou la réactivité du substrat (Figure 5A) (Lamontagne et al., 2003). Les quatre nucléotides de la boucle terminale forment la région appelée « Initial Binding and Positionning Box » ou IBPB. Tout changement de la taille de la boucle ou de la séquence des deux premières positions affecte grandement la liaison de l’enzyme et, par conséquent, abolit le clivage (Lamontagne et Abou Elela, 2004; Lamontagne et al., 2003). La seconde région, appelée « Binding Stability Box » ou BSB, est constituée des quatre paires de bases adjacentes à la boucle. Des changements de séquence dans cette région sont susceptibles d’affecter la stabilité du complexe enzyme:substrat. Par ailleurs, il a été montré que Rnt1p peut se lier stablement à 22

un ARN formé uniquement d’une tétra-boucle NGNN (IBPB) suivi de cinq paires de bases (BSB) (Lamontagne et al., 2003). Finalement, la troisième région dénommée « Cleavage Efficiency Box », ou CEB, regroupe les paires de bases entourant les sites de coupure. La présence de paires G-C à cet endroit influence le clivage, sans pour autant affecter la liaison de Rnt1p, ce qui suggère que la stabilité de l’hélice d’ARN à cet endroit module la réactivité du substrat (Lamontagne et al., 2003). Similairement, une étude publiée en 2011 a montré que la modification des séquences et des structures au niveau du CEB pouvait grandement moduler le taux de clivage in vivo et in vitro (Babiskin et Smolke, 2011b). De manière globale, on comprendra que la structure entourant la tétra- boucle sert de point initial à la reconnaissance de l’enzyme (IBPB+BSB), qui par la suite, permettra le positionnement du site catalytique (CEB).

Figure 5 : Les RNAse III reconnaissent différentes structures d’ARNdb ayant peu de conservation de séquence. (A) Rnt1p coupe l’ARN à 14 et 16 nucléotides (nt) d’une structure tige-boucle de type NGNN. Les flèches représentent les sites de coupure. Les trois régions encadrées (IBPB, BSB et CEB) représentent les épitopes qui affectent la liaison et le clivage de Rnt1p. Voir le texte principal pour plus de détails. (B) Comparaison des structures d’ARNdb reconnues par les autres classes de RNase III. Les RNase III procaryotes de la classe I reconnaissent une hélice d’ARNdb, dont la présence de séquences particulières (anti-déterminants) à l’intérieur des boîtes proximales (P), médianes (M) et distales (D) affecte l’activité et le positionnement des sites de clivage (Nicholson, 2014). Les substrats de Drosha (classe III) présentent une structure tige-boucle dont la taille et la séquence de la boucle ont peu d’influence sur la réactivité. La liaison des substrats et le positionnement de Drosha serait principalement dépendant du recrutement de la protéine partenaire DGCR8 au niveau de la base de la tige (Han et al., 2006). Finalement, la reconnaissance des substrats de Dicer repose principalement sur la liaison d’une extrémité saillante de 2 nucléotides en 3’ par le domaine PAZ. Le positionnement des sites de clivage sera déterminé en fonction de la distance entre les domaines PAZ et RIIID de Dicer (MacRae et al., 2006). 24

4.2 Le dsRDB de Rnt1p reconnaît la structure adoptée par la tétra-boucle

L’analyse par spectroscopie de résonance magnétique nucléaire des tige- boucles AGAA, AGUC et AGUU a montré que les tétra-boucles NGNN adoptent une structure relativement stable et distincte de celles de type GNRA ou UNCG (Lebars et al., 2001; Thapar et al., 2014; Wu et al., 2001). Les bases aux positions 1 et 2 de la boucle NGNN forment un empilement avec la base en 5’ qui ferme la boucle, et pareillement du coté 3’ avec les bases aux positions 3 et 4 (Wu et al., 2001). Le « G » conservé à la deuxième position adopte une conformation syn qui rend la base exposée à l’extérieur de la structure (Lebars et al., 2001). Cela induit aussi une torsion dans le squelette de l’ARN qui fait en sorte que la tétra-boucle adopte un repliement comparable à un sillon mineur qui est ensuite reconnu par le dsRBD (Figure 6) (Wu et al., 2004; Wu et al., 2001).

Figure 6 : Le dsRBD de Rnt1p reconnaît la structure de la tétra-boucle NGNN. Structure en solution du dsRBD de Rnt1p (magenta) en complexe avec une tétra- boucle AGAA (en orange et vert). Les quatre nucléotides formant la tétra-boucle sont numérotés. Notons la conformation syn de la guanosine conservée (position 2). Figure adaptée du modèle publié par Wu et al., 2004 (ID :1T4L du « RCSB Protein Data Bank; http://www.rcsb.org/pdb/home/home.do).

Le dsRBD a rapidement été identifié comme étant un déterminant majeur de l’activité et de la spécificité de Rnt1p. Sa présence au sein de la protéine est requise pour la maturation de l’ARNr in vivo et pour la liaison et le clivage in vitro (Lamontagne et al., 2000; Nagel et Ares, 2000). Des études structurelles du dsRBD de Rnt1p seul ou en complexe avec l’ARN montrent que l’hélice α1, l’hélice α2 et la boucle β1-β2 contactent respectivement les sillons mineurs, majeurs et 25

mineurs de l’ARN (Figure 6) (Thapar et al., 2014; Wu et al., 2004). Le dsRBD de Rnt1p possède en plus une hélice supplémentaire, α3, dont le rôle serait de faciliter la rotation et le positionnement de l’hélice α1 au niveau du sillon mineur de la tétra-boucle. Selon ce modèle, la conservation du « G » dans les tétra-boucles coupées par Rnt1p s’expliquerait donc par l’adoption d’une structure unique reconnue par le dsRBD particulier de Rnt1p (Wu et al., 2004). La spécificité de la liaison serait assurée par des changements conformationnels du dsRBD suite à la liaison d’ARNs apparentés ou non à une tétra-boucle NGNN (Hartman et al., 2013).

Malgré tout, plusieurs questions relatives à la reconnaissance spécifique des substrats par Rnt1p demeurent. Par exemple, quel est le rôle exact de la guanine conservée de la tétra-boucle? Des essais d’empreinte aux radicaux hydroxyl et de résonance magnétique nucléaire suggèrent que cette position dans l’ARN est effectivement contactée par la protéine in vitro, possiblement via un autre domaine de Rnt1p (Ghazal et Abou Elela, 2006; Lamontagne et al., 2003). Deuxièmement, quel rôle le NTD joue-t-il dans la reconnaissance des substrats? Il a été montré que le NTD est requis pour la maturation du pré-ARNr in vivo et pour assurer la stabilité du complexe enzyme:substrat in vitro (Ghazal et Abou Elela, 2006; Lamontagne et al., 2000). Troisièmement, comment le site de coupure est-il choisi en rapport avec la position de la tétra-boucle? À l’instar de la RNase III bactérienne, la coupure par Rnt1p n’est pas effectuée à proximité du site de liaison du dsRBD (Gan et al., 2006). Finalement, comment Rnt1p reconnaît-il la tétra- boucle AAGU (snR38) dont la structure tertiaire semble pourtant distincte de la tétra-boucle NGNN (Gaudin et al., 2006)? Une possibilité évoquée est que Rnt1p sélectionne différentes structures d’ARN en utilisant un mode alternatif de reconnaissance (Ghazal et Abou Elela, 2006). À l’opposé, une étude récente suggère que la tétra-boucle AAGU adopte une conformation similaire à la tétra- boucle AGAA suite à la liaison par le dsBRD de Rnt1p (Wang et al., 2011).

5. Objectifs du projet

Le développement normal des cellules ainsi que l’exposition à différents stress induit des changements importants et rapides de leur transcriptome. À ce titre, la dégradation de l’ARN joue un rôle important dans la régulation de l’expression des gènes en contrôlant à la fois la qualité et la quantité de l’ARN disponible dans une cellule. Malheureusement, les principes qui dictent la dégradation sélective de l’ARN ainsi que la contribution exacte des différentes ribonucléases dans la régulation de l’expression des gènes sont encore largement inconnus.

Les ribonucléases III représentent une famille d’enzymes ubiquitaires qui influencent l’expression des gènes à de nombreux niveaux. Par exemple, chez la levure Saccaromyces cerevisiae, l’expression de Rnt1p est essentielle pour la maturation des ARNnc et l’expression adéquate de nombreux ARNm. Toutefois l’impact direct de Rnt1p sur la stabilité de la majorité de ces gènes n’a pas été clairement établi. De plus, les mécanismes qui dictent la reconnaissance spécifiques des tétra-boucles NGNN par Rnt1p ne sont pas entièrement élucidés. Par conséquent, le but principal de mon projet était de caractériser le mécanisme d’action de Rnt1p afin de comprendre le rôle de cette ribonucléase dans la régulation de l’expression des gènes chez la levure.

La compréhension des mécanismes de reconnaissance des substrats est un prérequis obligatoire pour la caractérisation des fonctions biologiques et de la régulation de Rnt1p. Des études récentes démontrent que la reconnaissance des substats par Rnt1p n’est pas strictement limitée à la structure tige-boucle NGNN, mais aussi à une tétra-boucle AAGU ayant une strucutre très différente. Afin de mieux définir ce qui constitue un substrat de Rnt1p et comment celui-ci discrimine entre les différentes structures d’ARN double-brin, j’ai entrepris de caractériser les déterminants biochimiques et structuraux requis pour la reconnaissance spécifique et le clivage. Ces études devraient permettre d’identifier les étapes limitantes pour 27

la liaison et la coupure des ARNs et ainsi mieux comprendre les règles qui dictent la sélection des substrats par la RNase III.

Alors que la délétion des ribonucléases modifie l’abondance de nombreux ARNm, il n’est pas clair dans quelle proportion de ces changements sont directement causés par un changement de leur stabilité. Je vais donc utiliser une combinaison de méthodes génétiques et génomiques afin d’identifier l’ensemble des substrats directs de Rnt1p et mesurer l’impact de sa délétion sur le transcriptome. L’établissement du répertoire des substrats de Rnt1p devrait apporter une meilleure compréhension du rôle global de Rnt1p et de la dégradation nucléaire dans la régulation de l’expression de gènes. À partir des résultats obtenus, je vais par la suite chercher à comprendre comment le clivage par Rnt1p est lui-même régulé en changeant le contexte biologique dans lequel sont placées les signaux de dégradation ou bien en changeant les conditions de culture des levures. Au final, ces expériences devraient mettre en lumière des nouveaux mécanismes de régulation de l’ARN double-brin chez la levure tout en soulignant l’importance de la dégradation sélective de l’ARN pour la modulation de l’expression génique. 28

CHAPITRE I

La ribonucléase III Rnt1p de Saccharomyces cerevisiae utilise un réseau de ponts hydrogènes pour lier et cliver ses substrats.

AVANT PROPOS

Yeast Ribonuclease III Uses a Network of Multiple Hydrogen Bonds for RNA Binding and Cleavage.

Mathieu Lavoie et Sherif Abou Elela

Article publié dans

Biochemistry, volume 47, numéro 33, pages 8514-8526, 2008.

Contribution : J’ai réalisé toutes les expériences présentées dans cet article. De plus, j’ai préparé toutes les figures et rédigé les légendes des figures et la section « matériel et méthodes ». Sherif Abou Elela a rédigé le reste du manuscrit initial auquel j’ai apporté des modifications. J’ai contribué aussi à toutes les révisions subséquentes.

RÉSUMÉ

La RNase III de Saccharomyces cerevisiae, appelée Rnt1p, reconnaît et clive spécifiquement des ARNs double-brins coiffés d’une tétra-boucle NGNN. La reconnaissance de la structure particulière adoptée par l’ARN serait dépendante de la formation de plusieurs ponts hydrogènes entre Rnt1p et ses substrats. Dans le but d’évaluer la contribution de chaque pont hydrogène pour la liaison et le clivage par Rnt1p, une série de substrats comprenant une ou plusieurs substitutions des groupements 2’-hydroxyl par des 2’-O-methyl ou 2’-fluoro a été développée. Les mutations au site de clivage ont peu d’effet sur la liaison et le clivage indiquant que les groupements 2’-hydroxyl ne sont pas requis pour la catalyse. Alternativement, plusieurs substitutions au niveau de la tétra-boucle ont réduit le taux de clivage, suggérant que Rnt1p requiert la formation d’un réseau complexe de ponts hydrogènes avec le sommet de la structure d’ARN. Aucune des modifications 2’-fluoro individuelles n’a abolit le clivage, démontrant ainsi la redondance fonctionnelle des interactions. La position des groupements 2’- hydroxyl critiques pour la liaison et le clivage est dépendante de la structure et de la séquence de la tétra-boucle. Dans l’ensemble, les résultats indiquent que Rnt1p utilise deux modes de liaison. Le premier est principalement médié par le domaine de liaison à l’ARN double-brin et mène à la formation d’un complexe enzyme : substrat stable. De son côté, le second mode mène au clivage de l’ARN et requiert la présence des domaines nucléase et N-terminal.

ARTICLE 1 :

Yeast RNase III Uses a Network of Multiple Hydrogen Bonds for RNA Binding and Cleavage

Mathieu Lavoie and Sherif Abou Elela

Groupe ARN / RNA Group, Département de Microbiologie et d’Infectiologie, Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec, Canada J1H 5N4

Received February 9, 2008; Revised Manuscript Received June 20, 2008

Running title: Rnt1p uses multiple hydrogen bonds for RNA selection

Abbreviations: Rnt1p, Saccharomyces cerevisiae orthologue of bacterial RNase III; dsRNA, double- stranded RNA; NUCD, nuclease domain; dsRBD, double-stranded RNA binding domain; G2 and A1- loops, RNA stems capped with a NGNN or AAGU tetraloop, respectively; 2’-O-Me, 2’-O-methyl-ribonucleotide; 2’-F, 2’-fluoro- ribonucleotide; TL, target RNA; EMSA, electrophoretic mobility shift assay; dT, inverted deoxythymidine; ∆N-term, recombinant version of Rnt1p lacking the N- terminal extension; RBM, RNA binding motif.

Abstract

Members of the bacterial RNase III family recognize a variety of short structured RNAs with few common features. It is not clear how this group of enzymes supports high cleavage fidelity, while maintaining a broad base of substrates. Here we show that the yeast orthologue of RNase III (Rnt1p) uses a network of 2’-OH-dependent interactions to recognize substrates with different structures. We designed a series of bipartite substrates permitting the distinction between binding and cleavage defects. Each substrate was engineered to carry a single or multiple 2’-O-methyl or 2’-fluoro-ribonucleotides substitutions to prevent the formation of hydrogen bonds with a specific nucleotide or group of nucleotides. Interestingly, introduction of 2’-O-methyl-ribonucleotides near the cleavage site increased the rate of catalysis indicating that 2’-OH are not required for cleavage. Substitution of nucleotides in known Rnt1p binding site with 2’-O-methyl- ribonucleotides inhibited cleavage while single 2’-fluoro-ribonucleotides substitutions did not. This indicates that while no single 2’-OH is essential for Rnt1p cleavage, small changes in the substrate structure are not tolerated. Strikingly, several nucleotides substitution greatly increased the substrate dissociation constant with little or no effect on the Michaelis-Menten constant or rate of catalysis. Together, the results indicate that Rnt1p uses a network of nucleotides interactions to identify its substrate and support two distinct modes of binding. One mode is primarily mediated by the dsRNA binding domain and leads to the formation of stable RNA/protein complex, while the other requires the presence of the nuclease and N-terminal domains and leads to RNA cleavage.

Introduction

The bacterial RNase III family is comprised of enzymes that specifically cleave RNA duplexes (1, 2) to regulate gene expression (3), process non-coding RNA (4-6) and enforce cellular immunity (7). The RNase III family is divided into 4 classes, based on protein features (1). Class I includes bacterial enzymes that possess a single N-terminal nuclease domain (NUCD) and a C-terminal double stranded RNA binding domain (dsRBD) (8). Class II enzymes, which are found in fungi, possess in addition to the NUCD and the dsRBD a highly variable N-terminal domain with no apparent functional motifs. Class III enzymes contain two NUCDs and include plant and vertebrate enzymes like human Drosha (9). Class IV includes the RNAi enzyme Dicer, which possesses N-terminal helicase and PAZ domains (10). Yeast Rnt1p (11) is a class II enzyme harboring a dsRBD motif followed by a C-terminal extension required for nucleolar localization (12). The NUCD of Rnt1p exhibits the RNase III signature sequence implicated in catalysis. Rnt1p’s N-terminal extension promotes enzyme homodimerization and is required for efficient cleavage at high salt concentrations (13). Recently, crystal (14) and solution (15) structures of Rnt1p dsRBD confirmed the classical αβββα structure, and revealed an additional helix near the C-terminus (α3) that is unique to Rnt1p. The solution structure of the dsRBD / RNA complex (15) indicates that the additional helix is not located near the substrate RNA, but that it could influence the binding of RNA to α1 (14).

In bacteria, RNase III discriminates its substrates from other structured RNAs using antideterminant nucleotides and as such, it is the absence of certain substrate features and not their presence, that triggers RNA cleavage (16). In contrast, eukaryotic RNase III uses different mechanisms of substrate selectivity in which antideterminants play a minor role (17). For example, Rnt1p prefers substrates that exhibit an NGNN (G2) or AAGU (A1) tetraloop structure (18). This unique preference for G2 or A1 loops is likely due to differences in the protein structure, such as the additional helix (α3) at the end of Rnt1p dsRBD. The structure of the Rnt1p dsRBD / RNA complex (15) shows that the protein monomer 33

contacts both the RNA major and minor grooves, but, surprisingly, not the conserved G in the second position of the terminal tetraloop (17). Deletion of this tetraloop, or substitution of the universally conserved G, inhibit cleavage and reduces binding under physiological conditions (19). To explain the lack of interaction between the dsRBD and the conserved G, it was suggested that this nucleotide does not directly contribute to substrate recognition but instead confers a structural fold recognized by Rnt1p (20). However, Rnt1p fails to cleave RNA hairpins capped with an ACAA tetraloop, which have an overall conformation that is similar to G2-loops (21) and recognizes the A1-loops that adopts a different structure (22). It is not clear how Rnt1p distinguishes between RNAs with seemingly different structures and sequence from others that appear to be structurally similar to known substrates.

The crystal structure of the bacterial RNase III / RNA complex suggests that bacterial enzymes identify their substrates via a complex network of interactions where hydrogen bonds play an important role. The dsRBD of bacterial RNase III contacts the ribose 2’-OH while the nuclease domain forms hydrogen bonds with the phosphate backbone (23). Consistently, the solution structure of Rnt1p dsRBD in complex with a model substrate revealed six 2’-OH-mediated hydrogen bonds (15). Interestingly, four of these interactions are located in the substrate primary binding site formed by the tetraloop and adjacent four base pairs (19). However, the functional impact of these hydrogen bonds and their contribution to substrate selection and cleavage remain unclear.

In this study, we investigated the hydrogen bonds requirement for substrate recognition by Rnt1p using bipartite substrates that allow the separation of defects in binding and cleavage. Single and multiple substitutions of ribonucleotides with 2’-O-methyl-ribonucleotides (2’-O-Me) or 2’-fluoro- ribonucleotides (2’-F) were performed to identify hydrogen bonds critical for substrate binding and cleavage. The results suggest that while Rnt1p cleavage does not require hydrogen bonds mediated interactions near the cleavage site, it interacts with multiple nucleotides near the stem-loop region to determine the substrate identity. The position of 34

critical 2’-OH differs depending on the sequence and structure of the tetraloop. This indicates that RNase III broad spectrum of substrates is maintained, at least in part, through a flexible network of interactions that adapts to changes in the substrate structure while maintaining the proper position of the nuclease domain relative to the cleavage site.

Materials and Methods

Strains and Plasmids.

Yeast strains used in Figure 2 were grown and manipulated using standard procedures (24, 25). The strain YHM111-U2L2 was generated by transforming the YHM111 strain (MATa, trp1, ura3-52, ade2-101, his3, lys2, snr20::LYS2) (26) with pRS314/U2∆Stem/L2 construct as previously described (27).

In Vitro Enzymatic Assays.

Recombinant Rnt1p and ∆N-term were produced in bacteria and FLPC purified as described (28). RNA transcripts were generated by T7 RNA polymerase and gel purified as described previously (18). Chemically synthesized RNA oligoribonucleotides were purchased from Integrated DNA Technologies (Coralville, IA) or University Core DNA Services (Calgary, AB, Canada). RNA fragments were 5’-end-labeled using [γ-32P]ATP as previously described (19). Reconstitution of in trans substrates were obtained by mixing equimolar amounts of different guides (the sequences are indicated in each Figure) and target RNAs (TL; 5’- GAUAAAGCUAAUGUUUUGGAAUCUUCAAGAUUAU GGAG-3’) and 3 fmoles of 5’-end-labelled target RNA was added as a tracer to track multiple turnover reactions. Full time courses were performed for all substrates with the exception of those in Figures 4 and 5 were selected time points chosen based on the reciprocal full time course of the unmodified substrate were examined. For single turnover conditions, 6 pmoles of guide RNA was mixed with 3 fmoles of 5’- end-labelled target RNA. Cleavage reactions were performed by incubating 30-60 35

nM Rnt1p with the different substrates for 10 min at 30°C in 20 µl reaction buffer (30 mM Tris-HCl (pH 7.5), 5 mM spermidine, 0.1 mM DTT, 0.1 mM EDTA (pH 7.5), 10 mM MgCl2) supplemented with 150 mM KCl (Figures 1 and 3) or 75 mM KCl (in Figures 4 and 5). Changing the monovalent salt concentration in Figures 4 and 5 was necessary to allow a direct comparison between Rnt1p and the salt sensitive ∆N-term version (13). Cleavage products were separated on 20% denaturing PAGE and quantified as described (18). Size markers were generated by alkaline hydrolysis of 5’-end-labeled target RNA. Calculations were performed using GraphPad Prism 4.03 software (GraphPad Software, CA). All experiments were repeated at least three times.

Electrophoretic Mobility Shift Assay (EMSA).

Protein binding experiments were performed as previously described (17) with 3 fmoles of 5’-end-labelled RNA. For bipartite substrates, 3 fmoles of 5’-end- labelled guide RNA were incubated with 2 pmoles of unlabelled target RNA prior to the incubation with Rnt1p. The protein concentrations used in the assays ranged from 0.5 µM to 9 µM. Bands were quantified using an Instant Imager with associated software (Packard, Meriden, CT) and calculations were performed using GraphPad Prism 4.03 software (GraphPad Software, CA). All experiments were performed at least three times. A typical EMSA result is presented in Supplementary Figure 2.

RNA Electroporation in Yeast Living Cells.

Preparation of electrocompetent cells and electroporation was performed essentially as previously described (27) with minor modifications. Briefly, YHM111- U2L2 yeast cells were grown at 30°C in 500 ml YC-Trp and harvested in late log phase. Electrocompetent cells were obtained with successive washes of the cell pellet with ice-cold 1 M sorbitol. The final pellet was resuspended in 500 µl ice-cold 1 M sorbitol and 40 µl of this suspension was used per electroporation. For each transformation, 2 nmoles of guide RNA was used. The pulse was performed at 1.5 kV, 25 µF, and 200 Ω with the Bio-Rad MicroPulser (Bio-Rad, Richmond, CA). 36

Immediately after the pulse, cells were diluted with 1 ml of ice-cold 1 M sorbitol and transferred into a tube containing 4 ml of YC-trp media containing 12.5 µg thiolutine / ml to stop de novo transcription. After 20 min incubation at 30°C, cells were harvested and total RNA was extracted.

Primer Extension.

Primer extension analysis was performed essentially as described before (27). Briefly, 5 µg total RNA was incubated with 1 ng 5’-end-labelled primer (5’- GAGTATGCCGCCAATTAGTG-3’) for 2 min at 65°C followed by 30 min incubation at 37°C. After addition of the reaction mixture, samples were incubated for 40 min at 42°C. Reactions were stopped by addition of EDTA and successive treatments with RNase A and Proteinase K. The extended products were separated on a 6% denaturing PAGE and visualized by autoradiography. The same primer was used for the sequencing reactions performed using T7 Sequencing Kit from USB (USB Corporation, Cleveland, OH).

Results

The Presence of Phosphate and Not 2’-Hydroxyl Group near the Cleavage Site Is Required for Cleavage by Rnt1p.

Earlier studies have shown that the replacement of ribonucleotides 5’ to the scissile bond with deoxyribonucleotides or 2’-O-methyl ribonucleotides reduces, but does not inhibit, cleavage by Rnt1p indicating that the 2’-OH group adjacent to the scissile bond is not directly involved in the cleavage chemistry (29). In contrast, multiple deoxyribonucleotides substitutions at the 3’-end of the G2- tetraloop, which do not alter the cleavage site, reduced binding and prevented cleavage by Rnt1p (29). The interpretation of these results is difficult because it is not clear whether they arise from failure to form critical hydrogen bonds, sugar-dependent changes in structure, or intolerance to mixed DNA / RNA helix (29). In order to better define the chemical requirement for substrate recognition and the position of 2’-OH critical 37

for interaction with Rnt1p, we introduced single or multiple ribonucleotide modifications and followed their impact on the binding and cleavage of a bipartite substrate that allows the separation of binding and cleavage events. This recently developed bipartite substrate (EL11:TL) (27) is based on the sequence of Rnt1p cleavage signal found near the 3’-end of U5 snRNA (30). As shown in Figure 1A, EL11:TL is formed by annealing two complementary RNA fragments: the first acts as a guide and contains Rnt1p binding site (e.g. G2 tetraloop) and the other acts as a cleavable target. As expected, cleavage of the bipartite substrate (EL11:TL) obeys the kinetics of natural RNA substrates (27) (Supplementary Figure 1) and allows cleavage in a single site within the target sequence (Figure 1A and Table 1). Surprisingly, replacement of all ribonucleotides below the 5th position from 3’-end of the G2 tetraloop with 2’-O-Me (EL18-EM:TL and EL11-EM:TL) did not inhibit, but instead enhanced the cleavage of the target sequence (Figure 1A). Introduction of additional 2’-O-methyl group at the 5’ nucleotide of the guide sequence (EL11- EM+1:TL) did not have additional effects on cleavage. On the other hand, modification of ribonucleotides at both ends of the guide sequence (EL11-5’3’M:TL) increased cleavage (Figure 1A) and turnover rate by 5 times when compared to the unmodified substrate (Table 1). This increase in the kcat value led to considerable increase in cleavage efficiency suggesting that the inclusion of 2’-O-methyl group at the 3’-end of the guide sequence accelerates catalysis.

Table 1. Kinetic parameters of Rnt1p cleavage of bipartite substrates

k /K’ Substrate K’ (µM) k (min-1) K’ (µM) cat M d cat M (min-1•µM-1) EL11:TL 2.3 ± 0.2 5.2 ±0.3 1.7 ± 0.3 3.0 ± 0.6 EL18-EM:TL 1.3 ± 0.1 14.9 ± 0.8 2.4 ± 0.4 6.3 ± 1.0 EL11-EM:TL 4.4 ± 0.3 33.1 ± 2.0 3.2 ± 0.5 10.3 ± 1.7 EL11-EM+1:TL 2.9 ±0.3 18.1 ± 1.4 1.6 ± 0.4 11.1 ± 2.7 EL11-5'3'M:TL 2.6 ±0.3 24.0 ± 1.4 1.8 ± 0.3 13.7 ± 2.5 EL11-dT:TL 2.8 ±0.3 33.1 ± 2.1 3.3 ± 0.5 10.2 ± 1.8 EL11-ES:TL 0.41 ±0.02 1.7 ± 0.1 0.4 ± 0.1 3.9 ± 1.1 EL11-LoopM:TL >7 n/d n/d n/d EL11-StemM:TL >7 n/d n/d n/d EL11-5'StemM:TL 5.0 ± 0.5 2.5 ± 0.4 3.6 ± 1.0 0.7 ± 0.2 EL11-A6M:TL 1.7 ±0.2 1.8 ± 0.2 0.6 ± 0.3 2.8 ± 1.6 EL11-G7M:TL 4.8 ± 0.5 n/d n/d n/d EL11-C10M:TL >7 n/d n/d n/d α The initial rate of cleavage of substrates was measured as a function of substrate concentration. Thirty nanomolar recombinant Rnt1p was incubated with 0.2-12.8 µM bipartite substrates at 30°C in presence of 150 mM KCl. The best-fit curves to a Michaelis-Menten scheme were calculated using GraphPad Prism 4.03 software. The apparent dissociation constants (K’d) were determined by EMSA as described in Material and Methods. Values are derived from at least three independent experiments and standard error of the mean calculated. “n/d” indicates “not determined”. 39

Figure 1. Rnt1p cleavage of substrates with chemically modified stems. (A) Rnt1p binding and cleavage does not require 2’-hydroxyl groups opposite of the cleavage site. Guide RNAs containing 2’-O-methyl- ribonucleotides (boxed regions) were annealed to 5’-end-labelled target RNA (TL, shown in grey) and tested for cleavage with recombinant Rnt1p. (B) The introduction of 2’-O-Me enhances Rnt1p cleavage independent of substrate sequence and structure. A standard substrate formed in cis derived from the U5 snRNA 3’-end (30) and its 2’-O-Me modified version were tested for cleavage in the absence (-) or presence (+) of Rnt1p. (C) Introduction of phosphorothioate, but not an inverted deoxythymidine, inhibits in vitro cleavage by Rnt1p. A mix of target (TL) and guide RNAs containing an inverted deoxythymidine at its 3’-end (EL11-dT) or 40

containing phosphorothioate bonded ribonucleotides (EL11-ES, grey box) were tested for cleavage by Rnt1p. In each assay, substrate / protein ratio of 30:1 was incubated for 10 minutes at 30°C under physiological salt concentration (150 mM KCl). Uncleaved substrates (T or S) and cleavage products (P) were separated by denaturing PAGE and the bands were quantified. Cleavage rates relative to the non-modified substrate were calculated from at least three independent experiments and the average values are indicated below the gels with a maximum standard deviation of the mean of ± 0.10. The position of the RNA ladder is shown on the left. Arrowhead indicates the observed cleavage site. The dashed line symbolizes an extension of 17 ribonucleotides at the 3’-end of TL.

We next tested the effect of modifying the 3’ terminal ribonucleotide on the cleavage of conventional in cis substrates. As shown in Figure 1B, a single nucleotide modification at the 3’-end of a model substrate found at U5 snRNA 3’- end (U5L-Stem-3’OMe) was sufficient to increase cleavage by 60% when compared to unmodified RNA (U5L-Stem). Therefore, this increase in cleavage is not specific to the stem sequence or to the 2’-O-methyl moiety since it could also be achieved by other modifications like addition of an inverted deoxythymidine (dT) (EL11-dT:TL) (Figure 1C). Introduction of phosphorothioate linkage in the strand opposing the cleavage site (EL11-ES:TL) enhanced binding in the absence of Mg2+

(Table 1), while strongly inhibiting cleavage (Figure 1C) and reducing both kcat and

K’M without affecting the catalytic efficiency (kcat/K’M). This situation is characteristic of an unproductive binding of the enzyme to its substrate. This suggests that the phosphate backbone does not play an important role in the formation of the stable RNA / protein complex provided by the dsRBD but instead is required for the formation of the nuclease domain dependent catalytic complex. Alternatively, the observed reduction in the cleavage efficiency may be explained at least in part by the fact that phosphorothioate are not stereospecific. In theory, the modified substrate EL11-ES:TL may form 211 different isomers (2 isomers for each 11 phosphorothioate linkages) with variable reactivities (31-33). We conclude that Rnt1p does not require the presence of 2’-OH groups of the ribonucleotides opposing the cleavage site. 41

In order to test the effect of nucleotide modifications on Rnt1p cleavage in vivo, we transformed different guide RNAs (used in Figure 1) into yeast strains expressing a previously established U2 snRNA reporter gene (27) that carries at its 3’-end a sequence complementary to the guide RNAs (Figure 2A). The cleavage product generated at U2 snRNA 3’-end was detected by reverse transcription using a primer complementary to the sequence downstream of the target cleavage site. As shown in Figure 2B, no cleavage products were detected when tRNA (lane 5) was used as a template or in RNA extracted from mock-transformed cells (lane 6). Bands arising from premature stops of the primer extension (Un) were observed with all RNAs including those extracted from mock-transfected cells. In contrast, weak guide-dependent cleavage products were observed upon the transformation of either unmodified RNA (lane 7) or guide RNA carrying an inverted dT (lane 12). As predicted, strong cleavage was detected upon transformation of the different 2’- O-Me-modified guide RNAs (lanes 8-10). In the case of EL18-EM, bands corresponding to guide-dependent but cleavage-independent stop in primer extension (St) were observed (lane 9 and data not shown). Surprisingly, transformation of EL11-ES (lane 11), which contains phosphorothioate linkages, produced strong cleavage in vivo when compared to unmodified RNA. This could be probably explained by the increase in the stability of EL11-ES in cell extracts when compared with that of the unmodified EL11 (data not shown), which may compensate for the poor reactivity of EL11-ES in vitro. It is also possible that the effect of the phosphorothioate linkage is offset in vivo by cellular factors which can either favor the formation of catalytically competent RNA / protein complexes (34) or influence the configuration of the substrate’s phosphate backbone. Northern blot analysis using probes specific to the sequence of U2 snRNA confirmed the results obtained by primer extension and indicated that EL18-EM and EL11-5’3’M are the most effective guide RNAs in vivo (data not shown). The results confirm Rnt1p’s tolerance for the presence of 2’-O-methyl-ribonucleotide opposing the cleavage site and suggest that in yeast, as in mammalian cells (35), 2’-O-methyl-modified oligonucleotides are good tools for gene silencing. 42

Figure 2. Chemically modified guide RNAs enhance cleavage by Rnt1p in vivo. (A) Schematic representation of the U2 3’-end flanking region in which the canonical Rnt1p processing signal was replaced with a 36 bp fragment (L2) containing a sequence complementary to the 3’ extension of the guide RNAs (in grey). The position of the oligonucleotide used for primer extension (P.E.) is indicated. (B) Modified guide RNAs direct precise cleavage of U2 3’-end in vivo. Total RNA was extracted from yeast cells electroporated with water (Mock) as a negative control or with the different guide RNAs described in Figure 1. The cleavage site was mapped by reverse transcription using the 5’- end-labelled primer (P.E.) shown in (A). tRNA was used as template in the reverse transcription reaction to ensure probe specificity. The primer extension products were separated on 6% denaturing PAGE and compared to DNA sequence produced by the same primer. The arrowhead indicates the position of the guide- dependent cleavage site. Unspecific (Un) and secondary (St) products caused by cleavage-independent polymerase arrest are indicated on the right. 43

Introduction of 2’-O-methyl Groups within the G2 Tetraloop Inhibits Binding and Cleavage

The solution structure of Rnt1p dsRBD in complex with a model RNA substrate identified 2’-OH-dependent protein contacts with 4 variable nucleotides near the conserved G2 tetraloop. In contrast, no hydrogen bond interactions were identified with the highly conserved essential guanosine in position 2 of the tetraloop (36). In order to evaluate the importance of the 2’-hydroxyl moiety for substrate recognition, we introduced single and multiple 2’-O-Me near the G2 tetraloop and measured the impact on Rnt1p binding and cleavage. 2’-O-Me substitutions of all four nucleotides in the tetraloop (EL11- LoopM:TL), inhibited binding and cleavage (Figure 3 and Table 1) indicating that the ribose 2’-OH moieties in the tetraloop are important for both activities. Modification of the 5 bp below the tetraloop (EL11-StemM:TL) inhibited both binding and cleavage (Table 1 and Figure 3). Substitutions of the five ribonucleotides of one strand of the dsRNA helix at the 5’-end of the loop (EL11-5’StemM:TL) reduced binding and inhibited cleavage albeit not to the same extent as that observed upon the substitution of both strands of the helix (EL11-StemM:TL). These results demonstrate the importance of the upper stem 2’-OH moieties for Rnt1p binding and cleavage. Substitution of the adenosine in the first position of the loop with 2’-O-Me (EL11-

A6M:TL), which does not hydrogen bond with the dsRBD (15), reduced both K’M and kcat but had no effect on catalytic efficiency or binding (Table 1 and Figure 3). Single 2’-O-Me substitution of either the conserved guanosine (EL11-G7M:TL) or the first nucleotide downstream of the loop 3’-end (EL11-C10M:TL) affected binding and prevented cleavage. This strongly suggests that the 2’-OH of these two ribonucleotides support critical protein / RNA interactions in the context of the full enzyme, which cannot be detected in dsRBD / RNA complex (15). However, we cannot exclude at this stage the possibility that some or even all of the observed effects of the 2’-O-Me substitutions are caused by perturbation of the RNA structure or steric hindrance of Rnt1p binding. 44

Figure 3. Introduction of 2’-O-Me in Rnt1p primary binding site inhibits both binding and cleavage. Bipartite substrates containing 2’- O-methyl-ribonucleotides (boxed regions) were tested for cleavage by recombinant Rnt1p. In vitro cleavage reactions were performed essentially as described in Figure 1 with the exception that single turnover conditions were used. The dashed line indicates 17 ribonucleotides extension at the target 3’-end and the arrowhead indicates the position of the cleavage site. The position of substrates (T) and cleavage products (P) are indicated on the right, while the position of the RNA ladder is on the left. Cleavage rates relative to the non-modified substrate were calculated from three independent experiments and average values are indicated at bottom with a maximum standard deviation of the mean of ± 0.10.

2’-F Substitutions Identify Multiple 2’-OH-Dependent Interactions Required for Substrate Recognition.

In order to differentiate between ribonucleotides involved in the formation of hydrogen bonds critical for Rnt1p binding and cleavage from those essential for the formation of the RNA fold recognized by the enzyme, we individually substituted every ribonucleotide in or near the G2 tetraloop with 2’-fluoro ribonucleotide (Figure 4A). The 2’-F modification does not alter the conformation of the native ribonucleotide and thus minimizes the perturbation of the substrate structure. Moreover, the 2’-fluoro moiety exhibits a similar electronegativity pattern and bond polarity as the 2’-hydroxyl (37). As shown in Figure 4B and D, introduction of 2’-F in 45

the 5’-end (EL11-G5F:TL) or in the third position (EL11- U8F:TL) of the tetraloop had little negative effect on RNA cleavage. Changes in the first position of the tetraloop (EL11-A6F:TL) or in the second base-pair at the 3’-end of the loop (EL11- A11F:TL) reduced RNA cleavage by a modest 15%. The strongest effect (~30%) was observed by the substitution in the second (EL11-G7F:TL) or fourth position (EL11-C9F:TL) of the tetraloop or the closing base-pair at the 3’-end of the tetraloop (EL11-C10F:TL). Further analysis of the substitutions that affected the cleavage (EL11-A6F:TL, EL11-G7F:TL and EL11-C10F:TL) indicated that 2’-F substitutions in the second position of the tetraloop or at the 3’-end closing base- pair reduced binding affinity, while substitution in the first position of the loop has no effect on binding (Table 2). On the other hand, all three mutations reduced both

K’M and kcat with little effect on the catalytic efficiency (Table 2). These results indicate that no single 2’-OH in the tetraloop is absolutely essential for the interaction with Rnt1p. They also suggest that substrate recognition is achieved by the formation of a network of hydrogen bonds that involves, at least in part, the ribonucleotides in the second and fourth position of the G2 tetraloop. In order to directly test this hypothesis, we introduced multiple 2’-F substitutions in either position 1 and 2 of the tetraloop (EL11-A6G7F:TL) or position 4 and the first nucleotide of the stem (EL11-C9C10F:TL). As shown in Figure 4, multiple substitutions exerted synergistic inhibitory effect reducing the cleavage efficiency by more than 70% when compared with the unmodified substrate (EL11:TL). As expected, simultaneous substitution of the nucleotides at the 5’ end and 3’ end of the tetraloop (EL11-Combo4F:TL) almost completely inhibited the cleavage by Rnt1p. Together, the results suggest that Rnt1p cleavage efficiency is determined, at least in part, by the number of available hydrogen bonds in the upper stem-loop region.

Table 2. Kinetic parameters of Rnt1p or ∆N-term cleavage of 2’-fluoro- modified substrates

Rnt1p ∆N-Term kcat/K’M kcat/K’M K’d kcat K’M - kcat K’M -1 Substrate -1 (min -1 (min (µM) (min ) (µM) 1 -1 (min ) (µM) -1 •µM ) •µM ) EL11:TL 2.3 ± 0.2 4.2 ± 0.3 2.4 ± 0.3 1.7 ± 0.3 1.1 ± 0.1 0.5 ± 0.1 2.3 ± 0.3 EL11- 2.0 ± 0.1 1.7 ± 0.2 1.1 ± 0.3 1.5 ± 0.4 1.3 ± 0.1 0.7 ± 0.2 1.9 ± 0.6 A6F:TL EL11- 5.8 ± 0.4 1.6 ± 0.2 1.2 ± 0.3 1.4 ± 0.4 0.7 ± 0.1 0.5 ± 0.1 1.3 ± 0.3 G7F:TL EL11- 4.7 ± 0.3 1.0 ± 0.1 1.0 ± 0.2 1.0 ± 0.2 0.7 ± 0.1 1.4 ± 0.3 0.5 ± 0.1 C10F:TL EL11- 4.7 ± 0.2 1.6 ± 0.1 1.5 ± 0.3 1.1 ± 0.2 0.6 ± 0.1 1.0 ± 0.3 0.6 ± 0.2 AAGU:TL EL11-AAGU- 0.69 ± 0.51 ± 6.6 ± 0.5 1.9 ± 0.3 0.4 ± 0.1 1.8 ± 0.3 0.3 ± 0.1 A6F:TL 0.04 0.04 EL11-AAGU- 0.80 ± 0.44 ± 6.2 ± 0.3 1.0 ± 0.1 0.8 ± 0.1 1.2 ± 0.3 0.4 ± 0.1 A7F:TL 0.04 0.04 EL11-AAGU- 0.47 ± 0.25 ± 0.47 ± 5.0 ± 0.5 1.9 ± 0.3 1.1 ± 0.3 0.4 ± 0.1 C10F:TL 0.04 0.04 0.04 α The initial rate of cleavage of substrates was measured as a function of substrate concentration. Thirty nanomolar recombinant Rnt1p was incubated with 0.1-6.4 µM bipartite substrates at 30°C in presence of 75 mM KCl. Calculations were performed as described in Table 1.

Figure 4. Rnt1p cleavage of RNA capped with 2’-F modified G2 tetraloop. (A) Schematic representation of 2’-F (grey circles) modified guide RNAs annealed to their target (TL). The bipartite substrates were tested for cleavage by Rnt1p (in (B)) or by ∆N-term (in (C)), a protein 48

lacking the N-terminal domain of Rnt1p. Cleavage was carried essentially as described in Figure 1 except that the reactions were performed under 75 mM KCl with a substrate / protein ratio of 10:1. Three independent experiments were used to calculate the relative cleavage velocity (shown below the gels). The asterisk indicates the products generated by cleavage at a non-canonical site 13 nucleotides from the tetraloop. This aberrant cleavage product was included in the relative velocity calculation. (D) Graphical comparison of the cleavage rates obtained with different guide RNA modifications shown in (B) and (C). Rnt1p and ∆N-term values are shown in grey and white, respectively. Error bars represent the standard deviation of the mean.

The 2’-OH substitutions confirmed the presence of two of the hydrogen bonds predicted by the solution structure of Rnt1p dsRBD / RNA complex (C9 and A11) and identified two new potential hydrogen bonds (G7 and C10) that influence the cleavage reactions but could not be detected in the solution structure (Figure 4) (15). This suggests that the 2’-OH of G7 and C10 influence dsRBD-independent functions of Rnt1p. In order to test this possibility, we examined the impact of 2’- OH substitution on the binding and cleavage of a recombinant version of Rnt1p lacking the N-terminal domain (∆N-term) (13). The N-terminal domain is suggested to influence Rnt1p interaction with the 5’-end of the loop and stabilize the interaction of Rnt1p with its substrate (13). Therefore, we expect that substitutions of 2’-OH which contributes to N-terminal domain-dependent functions, to have no additional effect on ∆N-term activity, while substitutions acting on other domains should synergistically increase the defects caused by the deletion of Rnt1p N- terminal domain. As expected, deletion of the N-terminal domain did not affect the cleavage of the unmodified EL11:TL under low salt conditions, nor the modified substrates EL11-G5F:TL, EL11-A6F:TL and EL11-U8F:TL (Figures 4C and D). Modification in the second base-pair at the 3’-end of the loop (EL11-A11F:TL) or the conserved guanosine in the second position of the tetraloop (EL11-G7F:TL) modestly reduced the cleavage rate when compared to EL11:TL. Cleavage of EL11-C9F:TL and EL11-C10F:TL, in which the last position of the tetraloop or the first position in the stem below the tetraloop were modified, was strongly inhibited (Figure 4C). The ∆N-term cleavage kinetics of EL11-A6F:TL was very similar to 49

that of the unmodified RNA EL11:TL, while EL11-G7F:TL and EL11-C10F:TL reduced the rate of catalysis and the catalytic efficiency (Table 2) as observed with cleavage performed using the full enzyme. Deletion of the N-terminal domain had little additional effects on the cleavage of substrates carrying multiple 2’-F substitutions. This is probably due to the already the severe defects observed with the substrates carrying multiple substitutions. Nevertheless, analysis of the individual nucleotide substitutions indicate that Rnt1p N-terminal domain does not contribute to the enzyme interaction with C9 and C10 but influences, at least partially, 2’-OH-dependent interactions with G7 and A6. This is consistent with previous chemical footprinting and mutational analyses (18) suggesting that while the dsRBD is sufficient for binding with the 3’-end of the loop, the interaction of the enzyme with the 5’-end of the loop is influenced by the N-terminal domain. In addition, deletion of the N-terminal domain induced secondary cleavage at a non- canonical site 13 nucleotides from the tetraloop (Figure 4C) suggesting that the N- terminal domain ensures the correct positioning of the enzyme relative to its substrate.

Rnt1p Uses a Functionally Flexible Network of Hydrogen Bonds To Accommodate Changes in the Substrate Features.

In addition to the G2 tetraloop substrates, Rnt1p recognizes a class of RNA hairpins capped with AAGU tetraloop (A1 tetraloop), which differs in sequence and structure from the G2 class. Moreover, chemical footprinting experiments suggest that Rnt1p is positioned differently on A1 and G2-tetraloop substrates (18). It is not clear whether Rnt1p uses the same group of 2’-OH to identify the two classes of substrates or uses a substrate-specific network of interactions. In order to answer this question, we measured the effect of 2’-F substitutions in guide RNAs that share the sequence of EL11:TL with the exception of the tetraloop, which was changed from AGUC to AAGU (EL11-AAGU:TL) (Figure 5A). These substrates were evaluated for Rnt1p and ∆N-term binding and cleavage. As shown in Figures 50

5B and D, the unmodified A1-guide RNA (EL11-AAGU:TL) cleaved the target sequence as efficiently as observed with G2-guide RNA, EL11:TL (Figures 4B and D). However, unlike EL11:TL, the cleavage of EL11-AAGU:TL by the ∆N-term was significantly reduced (Figures 5C and D) indicating that A1 substrates are more dependent on the N-terminal domain than G2 substrates as previously suggested (18). Modification of the ribonucleotide at the 5’-end of the closing base-pairs (EL11-AAGU-G5F:TL) did not negatively affect Rnt1p or ∆N-term cleavage. All other modifications reduced the relative cleavage rate of Rnt1p by more than 30% with the most striking effect obtained by the modification of the nucleotide below the 3’-end of the tetraloop (EL11-AAGU-C10F:TL). This result is consistent with a cooperative binding mode of A1 substrates in which more nucleotides contribute to substrate recognition than observed in the case of G2 substrates (18). Interestingly, the modification of C10 also reduced the cleavage specificity resulting in a strong cleavage at a secondary site closer to the loop underscoring the contribution of this nucleotide for the positioning of the enzyme on its substrate. Multiple 2’-F substitutions (EL11-AAGU-A6A7F:TL, EL11-AAGU-U9C10F:TL and EL11-AAGU-Combo4F:TL) exerted additive inhibitory effects, preventing cleavage at the canonical cleavage site, while enhancing the cleavage at an alternative site closer to the tetraloop (Figure 5). Additional reduction in the cleavage rate by the ∆N-term was observed with all substrates carrying single nucleotide modifications with the exception of EL11-AAGU-G5F:TL and EL11-AAGU-A7F:TL (Figures 5C and D). Deletion of the N-terminal domain did not affect the cleavage of EL11- AAGU-A6A7F:TL, EL11-AAGU-U9C10F:TL or EL11-AAGU-Combo4F:TL, but instead enhanced the cleavage at the alternative cleavage site observed upon the modifications of the A1-tetraloop (Figure 5). Unlike their G2 counterparts, the substitution of either the first or the second position of the tetraloop reduced enzyme binding, lowered the rate of catalysis and decreased the catalytic efficiency (Table 2). Modification of the 3’-end nucleotide of the tetraloop closing base-pair did not affect Rnt1p binding or K’M but reduced the rate of catalysis and catalytic efficiency. In general, while all tested modifications in the G2 tetraloop reduced both K’M and kcat, the modifications in the A1 substrate reduced the kcat 51

without affecting the K’M exposing the differences in the role played by the 2’- hydroxyl groups of these tetraloops for Rnt1p cleavage. We conclude that Rnt1p uses different patterns of 2’-OH-dependent interactions for the recognition of A1 and G2 tetraloops

Figure 5. Rnt1p cleavage of RNA capped with 2’-F modified A1 tetraloop. (A) Illustration of the 2’-F (grey circles) modified A1-guides annealed to their targets (TL). The different substrates were cleaved by Rnt1p (B) or ∆N-term (C), the relative cleavage rates calculated and graphically compared (D) as described in Figure 4. 53

Discussion In this study we have shown that Rnt1p, the yeast orthologue of the bacterial RNase III, does not require 2’-hydroxyl groups near the cleavage site but instead requires the presence of the RNA backbone phosphates opposing the cleavage site. This indicates that the formation of the catalytic core requires, at least in part, interaction with the phosphate backbone explaining the broad specificity of the nuclease domain. Indeed, the structure of Aquifex aeolicus RNase III / RNA complex revealed that the nuclease domain RNA binding motif (RBM) 3, which is conserved in Rnt1p, forms a hydrogen bond with the non-bridging phosphate oxygen (23). In addition, crystal structures of the bacterial RNase III revealed that the nucleophilic attack occurs through coordination of metal ions and water molecules with the phosphate at the scissile bond (38). The ribose 2’-OH groups do not seem to have a major contribution in the phosphoryl transfer reaction (38). In contrast, the enzyme requires the presence of multiple 2’-OH groups within the substrate binding site (upper stem and tetraloop) for cleavage. The location and number of required 2’-OH differ depending on the sequence and structure of the substrate tetraloop suggesting that a flexible interactions network allows the enzyme to recognize different tetraloop folds (e.g. both A1 and G2-substrates). Together, the data suggest that Rnt1p may form two different RNA / protein complexes: the first is a magnesium-independent non-productive complex that forms through 2’-OH-dependent interactions between the dsRBD and the upper stem-loop, while the second complex leads to cleavage and involves the formation of phosphate-dependent interactions with the nuclease domain.

Rnt1p Requires Different Chemical and Structural Features for Substrate Binding and Cleavage.

Like bacterial RNase III (39), Rnt1p can form a stable complex with RNA in the absence of Mg2+ (13) and binds small RNA hairpins shorter than 11 bp (20). In addition, mutations in the protein or the RNA substrate that decrease binding 54

affinity in the absence of Mg2+ do not always lead to cleavage inhibition and several mutations that prevent catalysis do not inhibit binding (17). This suggests that Rnt1p may form two different complexes with different substrate requirements. The first is a non-productive complex that requires recognition by the dsRBD and forms independently of Mg2+. The second is Mg2+-dependent and requires interaction with both the dsRBD and the nuclease domain and lead to cleavage (17). Consistently, we have found that 2’-O-Me substitution of ribose 2’-OH in the strand opposing the cleavage site increases the rate of catalysis (kcat) by up to 5 2+ times with little effect on the Mg independent substrate affinity (K’d) and without 2+ obligatory change in the apparent Michaelis constant (K’M) in the presence of Mg (Table 1). Evidence for the presence of non-productive binding during the cleavage reaction could also be deduced from the fact that several individual 2’-fluoro modifications of ribonucleotides forming the primary binding site (upper stem-loop) decreased both kcat and K’M without impact on the catalytic efficiency (Table 2).

Once again, no strict correlation between 2’-OH substitutions that reduce the K’M and those that increase the K’d could be observed suggesting that the complex formed under non-catalytic conditions and the one forming prior to catalysis use different substrate requirements. Indeed, crystal structure of bacterial RNase III carrying mutation in the catalytic domain revealed that the dsRBD domain could independently bind to RNA outside the nuclease domain (23). It was also demonstrated that Mg2+ is required to induce the conformational change required for the positioning of the RNA duplex within the catalytic valley (23), stabilization of the catalytic assembly and product release (38). Therefore, we conclude that binding and cleavage of RNA duplexes by Rnt1p, like other RNase IIIs, are achieved through several transitional steps requiring different chemical interactions and structural conformation. Formation of productive or non-productive complexes, as well as transition between them, likely depends on the substrate’s ability to fulfill these requirements.

Distinguishing between Structural and Chemical Requirements for Rnt1p Substrate Reactivity.

To date, most studies of protein / RNA interactions used sugar substitutions such as 2’-deoxy, 2'- deoxy-2'-fluoro-β-D-arabinonucleosides (2’F-ANA) or 2’-O- methyl to identify hydrogen bonds required for RNA identification. Unfortunately, these types of substitutions often disturbs local structure (40) and hinder protein interaction (41), which makes the identification of hydrogen bonds critical for substrate selection very difficult. Substitutions with 2’-fluoro ribonucleotides, which preserve the RNA structure and cause minimum disturbance of hydrogen bonds independent protein interaction, are good alternative for evaluating the role of the tetraloop ribonucleotides in Rnt1p binding and cleavage. Organic fluorine can act as a weak hydrogen bond acceptor under certain conditions that optimize bond formation (42). However, the strength of fluorine-based interactions and their functional relevance to enzyme activity remain unclear and largely depend on the strength of the donor group, neighboring atoms and distance between partners (42). In most cases, fluorine-substituted ligands exhibit lower binding affinity and enzymatic activity when the hydrogen bond formation is a prerequisite for enzyme binding and catalysis (43). Thus, fluorine substitutions are excellent for probing hydrogen bond requirements especially those believed to be essential for Rnt1p binding and cleavage as those examined in this study. For example, fluorine substitution is particularly suitable for examining the contribution of the 2’-OH of the highly conserved guanosine present in the second position of the G2 tetraloop to binding and cleavage. Substituting this guanosine with another ribonucleotide (19), with deoxyguanosine (29) or with 2’-O-methyl-guanosine (Figure 3) strongly reduced Rnt1p binding and inhibited cleavage despite the fact that no interactions were detected between this nucleotide and the dsRBD by NMR (15). This led to the conclusion that the conserved G is only important for the tetraloop fold recognized by Rnt1p (20) and does not contribute biochemically to substrate recognition. By substituting this nucleotide with 2’-fluoro-guanosine we were able to reveal that the 2’-hydroxyl group of this nucleotide is equally important for substrate 56

cleavage as the nucleotide at the fourth position of the tetraloop (Figure 6A), which was shown to hydrogen bond with the dsRBD in the solution structure of Rnt1p dsRBD / RNA complex (15). This underscores the advantage of 2’-fluoro modifications as a precise probe for 2’-OH requirements. More importantly, these results clearly show that the conserved G plays a dual role in preserving the G2- substrate structure and in forming direct and specific interaction with the enzyme. In general, the introduction of 2’-O-Me at any position near the enzyme primary binding site was much more inhibitory to binding and cleavage than given 2’-F substitutions (Tables 1 and 2). This indicates that Rnt1p is much more sensitive to changes in the overall structure of the tetraloop than to position-specific hydrogen bonding, which explains Rnt1p’s capacity to recognize tetraloops with different sequence composition (4).

Identification of Hydrogen Bonds Involved in Substrate Binding and Cleavage.

The solution structure of Rnt1p dsRBD in complex with a model G2 substrate indicated that the dsRBD forms a total of six direct or water-mediated hydrogen bonds with 2’-OH groups of the RNA (15). Four of those are formed between the upper stem-loop region and helix α1 suggesting that these interactions are critical for binding and perhaps for cleavage (Figure 6A). However, individual substitution of the 2’-OH involved in the formation of these four hydrogen bonds with 2’-F did not inhibit cleavage (Figure 6A). Substitutions that prevent hydrogen bond formation between the 5’-end of the closing base pair and the aspartic acid in position 367 of the dsRBD or between the third position of the tetraloop and arginine in position 372 did not have any negative effect on cleavage of G2 substrates (Figure 4). Together, these results suggest that not all hydrogen bonds formed between the dsRBD and the RNA are essential for cleavage. However, we cannot formally exclude the fact that some minor roles of certain hydrogen bonds observed in the NMR structure were missed because they were 57

maintained by the fluorine substitutions (42). On the other hand, it should also be noted that the hydrogen bonds observed in the solution structure between the RNA and the truncated dsRBD (15) may not reflect the interactions formed in the context of the full enzyme. Substituting the 2’-OH required for the formation of the hydrogen bond between the fourth nucleotide of the tetraloop and serine 376 moderately reduced cleavage. In addition, weak effect was observed upon the disruption of the hydrogen bond between the second nucleotide below the 3’-end of the tetraloop and lysine 371 (Figure 4).

Interestingly, some cleavage reduction was observed by substituting positions one and two of the tetraloop as well as the first nucleotide below the 3’- end of the tetraloop, which do not interact with the dsRBD in the NMR structure (15). This suggests that the dsRBD forms hydrogen bond with these nucleotides only in the context of the full enzyme. Another possibility is that different domains of Rnt1p interact with these 2’-OH groups. Consistently, positions 1 and 2 of the tetraloop seem to contribute to the function of the N-terminal domain (Figure 4D). This data is consistent with previous mutational analysis, chemical footprinting, and chemical interference that all indicate that the first and second positions interact with the enzyme and play an important role in cleavage (4, 17). The significant contribution of the first 2 nucleotides of G2-tetraloop to RNA cleavage explains in part their high conservation (4) and suggests that while Rnt1p may form complexes with a large number of RNA (18) and even DNA molecules (29), it can only cleave a subset that sponsors the fold of G2 or A1- tetraloops.

A Model for Rnt1p Substrate Identification and Cleavage.

Multiple crystal structures of bacterial RNase III / RNA complexes suggest that RNase III first binds its substrate through interactions with the dsRBD that lead, in the presence of Mg2+, to conformational changes in order to form the catalytic core (23). The catalytic core is formed through intermolecular homodimerization of the nuclease domains that fit the RNA substrate. In the 58

catalytic core, the RNA duplex is held in place by a total of four RBMs, which form extensive network of hydrogen bonds with two regions of the substrate surrounding the cleavage site (23). In addition, each enzyme subunit contains two other RBMs located in the dsRBD (Figure 6A). Since most RBMs, nuclease signature motif and secondary structure of catalytic domains are conserved in Rnt1p, we were able to use bacterial RNase III tertiary structure (23, 44, 45) as a model to position Rnt1p relative to its substrate based on the results of nucleotide substitution (29) (Figures 4 and 5), chemical interference (18), and the solution structure of Rnt1p dsRBD / RNA complex (15) (Figure 6A). The tertiary structure of Rnt1p NUCD was predicted via sequence alignment with the bacterial enzyme (data not shown; the method used to produce the tertiary model is described in the Supporting Information). In this model, the RBM 1 in the dsRBD interacts with the tetraloop upper stem region and is stabilized in place by intermolecular interaction between the dsRBD and the N-terminal domain (13). Rnt1p nuclease domain is positioned against the cleavage site in a similar fashion to that of bacterial RNase III and forms contacts with the phosphate backbone. The positioning of Rnt1p RBMs relative to the RNA was possible due to the natural bend of Rnt1p binding signal (15, 20), the slight increase in the size of the C-terminal helix of the NUCD and the linker region between the dsRBD and the NUCD (data not shown). The natural flexible linker of Rnt1p joined the dsRBD and NUCD without the need for structural modifications of the adjacent helices. The position of RBM 4 was more difficult to ascertain given its poor sequence and functional conservation (17, 23). The overall binding patterns of G2 and A1 substrates are very similar with the exception that the network of critical interactions in the A1 tetraloop area appears to be more evenly distributed between the 3’ and 5’-end of the tetraloop (Figure 6A), as previously observed with chemical footprinting experiments (18). In Figure 6B, we propose a model for Rnt1p binding and cleavage that explains why 2’-OH substitutions displayed position- dependent effects on Rnt1p activities. In this model, the primary interaction between the RNA and the dsRBD occurs independent of the nuclease domain and regardless of the presence or absence of the cleavage site. This interaction is stable in the absence of Mg2+ and is largely 59

2+ responsible for the K’d value of the substrate. In presence of Mg , the RNA will be placed within the catalytic valley if it contains a suitable cleavage site. Bacterial RNase III binds Mg2+ in the absence of RNA (45) and requires it to bind its substrate (38). However, it is unclear at which step the magnesium ions are recruited to catalytic core. The formation of the NUCD / RNA complex is mediated in part by phosphate backbone interactions while the overall stability of the enzyme / RNA complex is stabilized by interactions between the dsRBD and the N-terminal domains (13, 17-19). The catalytically poised complex also requires the interaction of the RNA with a set of amino acid residues in the dsRBD. This complex determines the enzyme K’M and kcat.

The new model of Rnt1p substrate interaction described here illustrates Rnt1p’s capacity to form alternative productive and non-productive RNA complexes and explains the molecular basis for the enzyme broad substrate specificity. Like RNase III, Rnt1p mechanism of action suggests that binding and cleavage occurs in at least two distinct steps requiring non-identical RNA features (23, 46). This reveals the potential capacity of these enzymes to influence RNA stability and function through the formation of cleavage-independent ribonucleoprotein complexes (47). In addition, the apparently distinct binding and nuclease domain docking and cleavage steps of RNase III illustrated here and proposed by other studies (44, 47) suggest that the enzyme may have evolved from the fusion of a dsRNA binding protein with a primitive endoribonuclease.

Figure 6. Hypothetical model for Rnt1p binding and cleavage. (A) Comparison between bacterial RNase III and Rnt1p mechanisms of substrate recognition. The position of the bacterial RNase III relative to its substrate was based on the co-crystal structure of the RNA protein complex (23). Rnt1p nuclease domain position relative to substrate was modeled based on sequence alignment with the bacterial RNase III and location of the cleavage site. The positioning of the dsRBD near the tetraloop was based on the solution structure of Rnt1p dsRBD / RNA complex (15), chemical footprinting experiments (18, 19) and nucleotide substitution (Figures 4 and 5). Models of G2 and A1 substrates in complex with Rnt1p or with a version of Rnt1p lacking the N-terminal domain (∆N- 61

term) are shown. The ribonucleotides are represented as rectangles and the arrows indicate the enzymes cleavage sites. The solid lines represent subunit 1 of the homodimeric enzyme, while hatched lines indicate subunit 2. The dsRNA binding, nuclease and N-terminal domains are shown in red, green and blue, respectively. The numbered circles represent the RNA binding motifs (RBMs) of the proteins identified based on the RNA / protein co-crystal structure of bacterial RNase III (23) and sequence comparison with Rnt1p. Reactivity epitopes on the substrates that are known to interact with the enzyme or affect its activity are boxed. Nucleotides shown to form hydrogen bonds with Rnt1p dsRBD via their ribose 2’-OH in RNA / dsRBD solution structure are outlined in orange (15). Open, grey and black circles indicate the position of previously observed chemical modifications causing respectively weak, moderate and strong interference with Rnt1p interaction (18). Position of 2’-F substitutions causing weak, moderate and strong reductions in cleavage (in Figures 4 and 5) are shown as hatched, grey and black rectangles, respectively. (B) Proposed mechanism for substrate binding and cleavage by Rnt1p. Like bacterial RNase III (23), Rnt1p forms an intermolecular homodimer (13) where subunits are likely held together through their nuclease domains. In the case of Rnt1p, intermolecular interaction between the dsRBD and the N-terminal domain was previously demonstrated both biochemically and genetically (13). The RNA substrate first interacts with the dsRBD leading to the formation of the catalytic complex in which the binding of the substrate to the nuclease domain triggers conformational changes in both the enzyme and the substrate allowing cleavage.

Acknowledgements

We thank Bruno Lamontagne for help with techniques and training of M.L. and the critical reading of the manuscript. We are grateful to Pierre Lavigne, Xinhua Ji and Jean-Pierre Perreault for stimulating discussions and comments on the manuscript.

References

1. Lamontagne, B., Larose, S., Boulanger, J., and Elela, S. A. (2001) The RNase III family: a conserved structure and expanding functions in eukaryotic dsRNA metabolism, Curr. Issues Mol. Biol. 3, 71-78. 2. Ritchie, W., Legendre, M., and Gautheret, D. (2007) RNA stem-loops: to be or not to be cleaved by RNAse III, RNA 13, 457-462. 3. Lykke-Andersen, K. (2006) Regulation of gene expression in mouse embryos and its embryonic cells through RNAi, Mol. Biotechnol. 34, 271- 278. 4. Ghazal, G., Ge, D., Gervais-Bird, J., Gagnon, J., and Abou Elela, S. (2005) Genome-wide prediction and analysis of yeast RNase III-dependent snoRNA processing signals, Mol. Cell. Biol. 25, 2981-2994. 5. Regnier, P., and Grunberg-Manago, M. (1990) RNase III cleavages in non- coding leaders of Escherichia coli transcripts control mRNA stability and genetic expression, Biochimie 72, 825- 834. 6. Saito, K., Ishizuka, A., Siomi, H., and Siomi, M. C. (2005) Processing of pre- microRNAs by the Dicer-1-Loquacious complex in Drosophila cells, PLoS Biol. 3, e235. 7. Gatignol, A., Laine, S., and Clerzius, G. (2005) Dual role of TRBP in HIV replication and RNA interference: viral diversion of a cellular pathway or evasion from antiviral immunity?, Retrovirology 2, 65. 8. Nicholson, A. W. (1999) Function, mechanism and regulation of bacterial ribonucleases, FEMS Microbiol. Rev. 23, 371-390. 9. Collins, R. E., and Cheng, X. (2005) Structural domains in RNAi, FEBS Lett. 579, 5841-5849. 10. Carmell, M. A., and Hannon, G. J. (2004) RNase III enzymes and the initiation of gene silencing, Nat. Struct. Mol. Biol. 11, 214-218. 11. Abou Elela, S., Igel, H., and Ares, M., Jr. (1996) RNase III cleaves eukaryotic preribosomal RNA at a U3 snoRNP-dependent site, Cell 85, 115- 124. 12. Catala, M., Lamontagne, B., Larose, S., Ghazal, G., and Abou Elela, S. (2004) Cell cycle- dependent nuclear localization of yeast RNase III is required for efficient cell division, Mol. Biol. Cell. 15, 3015-3030. 13. Lamontagne, B., Tremblay, A., and Abou Elela, S. (2000) The N-terminal domain that distinguishes yeast from bacterial RNase III contains a 63

dimerization signal required for efficient double-stranded RNA cleavage, Mol. Cell. Biol. 20, 1104-1115. 14. Leulliot, N., Quevillon-Cheruel, S., Graille, M., Van Tilbeurgh, H., Leeper, T. C., Godin, K. S., Edwards, T. E., Sigurdsson, S. T., Rozenkrants, N., Nagel, R. J., Ares, M., and Varani, G. (2004) A new alpha-helical extension promotes RNA binding by the dsRBD of Rnt1p RNAse III, EMBO J. 23, 2468-2477. 15. Wu, H., Henras, A., Chanfreau, G., and Feigon, J. (2004) Structural basis for recognition of the AGNN tetraloop RNA fold by the double-stranded RNA- binding domain of Rnt1p RNase III, Proc. Natl. Acad. Sci. U.S.A. 101, 8307- 8312. 16. Zhang, K., and Nicholson, A. W. (1997) Regulation of ribonuclease III processing by double-helical sequence antideterminants, Proc. Natl. Acad. Sci. U.S.A. 94, 13437-13441. 17. Lamontagne, B., and Abou Elela, S. (2004) Evaluation of the RNA determinants for bacterial and yeast RNase III binding and cleavage, J. Biol. Chem. 279, 2231-2241. 18. Ghazal, G., and Elela, S. A. (2006) Characterization of the reactivity determinants of a novel hairpin substrate of yeast RNase III, J. Mol. Biol. 363, 332-344. 19. Lamontagne, B., Ghazal, G., Lebars, I., Yoshizawa, S., Fourmy, D., and Abou Elela, S. (2003) Sequence dependence of substrate recognition and cleavage by yeast RNase III, J. Mol. Biol. 327, 985-1000. 20. Lebars, I., Lamontagne, B., Yoshizawa, S., Aboul-Elela, S., and Fourmy, D. (2001) Solution structure of conserved AGNN tetraloops: insights into Rnt1p RNA processing, EMBO J. 20, 7250-7258. 21. Staple, D. W., and Butcher, S. E. (2003) Solution structure of the HIV-1 frameshift inducing stem-loop RNA, Nucleic Acids Res. 31, 4326-4331. 22. Gaudin, C., Ghazal, G., Yoshizawa, S., Elela, S. A., and Fourmy, D. (2006) Structure of an AAGU tetraloop and its contribution to substrate selection by yeast RNase III, J. Mol. Biol. 363, 322-331. 23. Gan, J., Tropea, J. E., Austin, B. P., Court, D. L., Waugh, D. S., and Ji, X. (2006) Structural insight into the mechanism of double-stranded RNA processing by ribonuclease III, Cell 124, 355-366. 24. Guthrie, C., and Fink, G. R. (1991) Guide to Yeast Genetics and Molecular Biology, Academic Press, San Diego, CA. 64

25. Rose, M. D., Winston, F., and Hieter, P. (1990) Methods in Yeast Genetics: A Laboratory Course Manual, Cold Spring Harbor, New York. 26. Madhani, H. D., and Guthrie, C. (1992) A novel base-pairing interaction between U2 and U6 snRNAs suggests a mechanism for the catalytic activation of the spliceosome, Cell 71, 803-817. 27. Lamontagne, B., and Abou Elela, S. (2007) Short RNA guides cleavage by eukaryotic RNase III, PLoS ONE 2, e472. 28. Lamontagne, B., and Abou Elela, S. (2001) Purification and characterization of Saccharomyces cerevisiae Rnt1p nuclease, Methods Enzymol. 342, 159- 167. 29. Lamontagne, B., Hannoush, R. N., Damha, M. J., and Abou Elela, S. (2004) Molecular requirements for duplex recognition and cleavage by eukaryotic RNase III: discovery of an RNA-dependent DNA cleavage activity of yeast Rnt1p, J. Mol. Biol. 338, 401-418. 30. Chanfreau, G., Abou Elela, S., Ares, M., Jr., and Guthrie, C. (1997) Alternative 3'-end processing of U5 snRNA by RNase III, Genes Dev. 11, 2741-2751. 31. Burgers, P. M., and Eckstein, F. (1979) Diastereomers of 5'-O-adenosyl 3'- O-uridyl phosphorothioate: chemical synthesis and enzymatic properties, Biochemistry 18, 592-596. 32. Burgers, P. M., Sathyanarayana, B. K., Saenger, W., and Eckstein, F. (1979) Crystal and molecular structure of adenosine 5'-O-phosphorothioate O-p-nitrophenyl ester (Sp diastereomer). Substrate stereospecificity of snake venom phosphodiesterase, Eur. J. Biochem. 100, 585-591. 33. Yamanaka, G., Eckstein, F., and Stryer, L. (1985) Stereochemistry of the guanyl nucleotide binding site of transducin probed by phosphorothioate analogues of GTP and GDP, Biochemistry 24, 8094-8101. 34. Giorgi, C., Fatica, A., Nagel, R., and Bozzoni, I. (2001) Release of U18 snoRNA from its host intron requires interaction of Nop1p with the Rnt1p endonuclease, EMBO J. 20, 6856-6865. 35. Kraynack, B. A., and Baker, B. F. (2006) Small interfering RNAs containing full 2'-O- methylribonucleotide-modified sense strands display Argonaute2/eIF2C2-dependent activity, RNA 12, 163-176. 36. Henras, A. K., Sam, M., Hiley, S. L., Wu, H., Hughes, T. R., Feigon, J., and Chanfreau, G. F. (2005) Biochemical and genomic analysis of substrate recognition by the double-stranded RNA binding domain of yeast RNase III, RNA 11, 1225-1237. 65

37. Egli, M., and Gryaznov, S. M. (2000) Synthetic oligonucleotides as RNA mimetics: 2'-modified RNAs and N3'-->P5' phosphoramidates, Cell. Mol. Life Sci. 57, 1440-1456. 38. Gan, J., Shaw, G., Tropea, J. E., Waugh, D. S., Court, D. L., and Ji, X. (2008) A stepwise model for double-stranded RNA processing by ribonuclease III, Mol. Microbiol. 67, 143-154. 39. Li, H., and Nicholson, A. W. (1996) Defining the enzyme binding domain of a ribonuclease III processing signal. Ethylation interference and hydroxyl radical footprinting using catalytically inactive RNase III mutants, EMBO J. 15, 1421-1433. 40. Ray, A. S., Schinazi, R. F., Murakami, E., Basavapathruni, A., Shi, J., Zorca, S. M., Chu, C. K., and Anderson, K. S. (2003) Probing the mechanistic consequences of 5-fluorine substitution on cytidine nucleotide analogue incorporation by HIV-1 reverse transcriptase, Antivir. Chem. Chemother. 14, 115-125. 41. Dertinger, D., Dale, T., and Uhlenbeck, O. C. (2001) Modifying the specificity of an RNA backbone contact, J. Mol. Biol. 314, 649-654. 42. Howard, J. A. K., Hoy, V. J., O'Hagan, D., and Smith, G. T. (1996) How good is fluorine as a hydrogen bond acceptor?, Tetrahedron 52, 12613- 12622. 43. Nidetzky, B., Mayr, P., Hadwiger, P., and Stutz, A. E. (1999) Binding energy and specificity in the catalytic mechanism of yeast aldose reductases, Biochem. J. 344 Pt 1, 101-107. 44. Blaszczyk, J., Gan, J., Tropea, J. E., Court, D. L., Waugh, D. S., and Ji, X. (2004) Noncatalytic assembly of ribonuclease III with double-stranded RNA, Structure (Camb) 12, 457-466. 45. Blaszczyk, J., Tropea, J. E., Bubunenko, M., Routzahn, K. M., Waugh, D. S., Court, D. L., and Ji, X. (2001) Crystallographic and modeling studies of RNase III suggest a mechanism for double-stranded RNA cleavage, Structure (Camb) 9, 1225-1236. 46. Sun, W., Jun, E., and Nicholson, A. W. (2001) Intrinsic double-stranded- RNA processing activity of Escherichia coli ribonuclease III lacking the dsRNA-binding domain, Biochemistry 40, 14976-14984. 47. Calin-Jageman, I., and Nicholson, A. W. (2003) RNA structure-dependent uncoupling of substrate recognition and cleavage by Escherichia coli ribonuclease III, Nucleic Acids Res. 31, 2381-2392.

Supporting Information for:

Yeast RNase III Uses a Network of Multiple Hydrogen Bonds for RNA Binding and Cleavage

Mathieu Lavoie and Sherif Abou Elela

Molecular Modeling.

Sequence alignment between Aquifex aeolicus RNase III (Protein Data Bank, 2EZ6) and Saccharomyces cerevisiae Rnt1p (Saccharomyces Genome Database, YMR239C) was performed using Insight II (2000) Suite software (Accelrys, CA) and secondary structures (not shown) for both full proteins were determined using the Jpred server (1). Using the Insight II suite, Rnt1p NUCD (from position Pro203 to Leu360) was folded using AaRNase III NUCD (from position Met1 to Val148) as a model. Folding was based on alignments of carbon- alpha’s coordinates. The insertions in Rnt1p sequence created an unstructured loop that could possibly interfere with the catalytic core. Therefore, residues Leu294 to Tyr315 were selected and a search was made in the Insight II suite’s database for the most probable loop that could be adopted by these residues without interfering with other residues or with the formation of the catalytic core or docking of the RNA substrate. The RNA presented in the solution structure of Rnt1p dsRBD / RNA complex (2) (Protein Data Bank, 1T4L) was extended with A- U base pairs to generate a 38 bp-stem. The folded Rnt1p NUCD was positioned on the RNA in order to place the conserved acidic residues forming the catalytic core within reasonable distance of the expected scissile bond. Torsion angles in the linker region were manually adjusted to connect the NUCD with the N-terminal dsRBD residues. The second subunit was created by duplicating the first NUCD / dsRBD complex and also positioned against the expected cleavage site. Positioning of the second subunit dsRBD was not completed since we have no direct evidence if it binds RNA and if so, where and how. The resulting tertiary structure was used to generate the schematic representation shown in Figure 6 (3). 67

Supporting Information References

1. Cuff, J. A., Clamp, M. E., Siddiqui, A. S., Finlay, M., and Barton, G. J. (1998) JPred: a consensus secondary structure prediction server, Bioinformatics 14, 892-893. 2. Wu, H., Henras, A., Chanfreau, G., and Feigon, J. (2004) Structural basis for recognition of the AGNN tetraloop RNA fold by the double-stranded RNA- binding domain of Rnt1p RNase III, Proc. Natl. Acad. Sci. U.S.A. 101, 8307- 8312. 3. Gan, J., Tropea, J. E., Austin, B. P., Court, D. L., Waugh, D. S., and Ji, X. (2006) Structural insight into the mechanism of double-stranded RNA processing by ribonuclease III, Cell 124, 355-366.

Supplementary Figures

Supplementary Figure 1. Rnt1p cleaves enzymatically- and chemically-synthesized substrates with similar reactivity. T7- synthesized substrate (CIS-EL11:TL) and chemically synthesized substrate (EL11:TL) were incubated in absence (-) or presence (+) of recombinant Rnt1p as described in Figure 1. Uncleaved substrates (T) and the cleavage products (P) were separated by denaturing PAGE and the bands were quantified. The cleavage rate relative to CIS-EL11:TL was calculated from at least three independent experiments and the average values are indicated below the gels. The maximum standard deviation of the mean is ± 0.10. The position of the RNA ladder is shown on the left. Arrowhead indicates the observed cleavage site. The kinetic parameters of substrates were calculated as detailed in Table 1. The apparent dissociation constants (K’d) were determined by EMSA as described in Material and Methods. Values are derived from at least three independent experiments and standard error of the mean calculated. 69

Supplementary Figure 2. An example of the mobility shift assay (EMSA) used to calculate the dissociation coefficient of the different Rnt1p substrates. As detailed in Materials and Methods, 5'-end labelled substrate (EL11:TL in this example) was incubated with increasing amount of recombinant Rnt1p (or ΔN-term) and separated on a 4% non- denaturating polyacrylamide gel. The different bands were displayed and directly quantified using an Instant-Imager (Packard, Meriden, CT), which provides accurate counts per minutes (CPM) for each band.

CHAPITRE II

Rnt1p utilise deux systèmes de mesure lors de la reconnaissance de ses substrats

AVANT PROPOS

Structure of a Eukaryotic RNase III Postcleavage Complex Reveals a Double- Ruler Mechanism for Substrate Selection

Yu-he Liang, Mathieu Lavoie, Marc-André Comeau, Sherif Abou Elela et Xinhua Ji

Article publié dans

Molecular Cell, volume 54, numéro 3, pages 431-444, 2014

Contribution : J’ai participé au design et complété les expériences présentées dans les Figures 3C-E, 4C, 5C-D, 6B-C et 7, ainsi que les figures supplémentaires S1A,D-E ET S3E-F. De plus, j’ai effectué les clonages et les purifications des mutants de la protéine Rnt1p utilisés dans l’ensemble de l’article. Yu-he Liang a réalisé la croissance des cristaux et déterminé la structure du complexe. Marc- André Comeau a effectué les essais de liaison et déterminé les constantes cinétiques. Au niveau de la rédaction, j’ai participé à la préparation et à l’assemblage des figures et rédigé les méthodes correspondantes aux expériences que j’ai réalisé (clonage, purification des protéines, essais de clivage in vitro). Xinhua Ji et Sherif Abou Elela ont rédigé l’ensemble du manuscrit initial auquel j’ai apporté des modifications. J’ai aussi contribué à toutes les révisions subséquentes. 71

RÉSUMÉ

La structure et le mécanisme d’action des ribonucleases III bactériennes sont bien caractérisés. Malheureusement, le modèle procaryote n’est pas suffisant pour expliquer le fonctionnement des RNase III eucaryotes, notamment la reconnaissance spécifique des substrats et le rôle des domaines additionnels à l’extrémité N-terminale des protéines. Nous avons donc déterminé la structure cristalline de Rnt1p, la RNase III de la levure Saccharomyces cerevisiae, en complexe avec son produit de clivage. Les résultats ont notamment permis de valider que deux acides aminés supplémentaires, propres aux enzymes eucaryotes, participent à la formation du centre catalytique. De manière similaire à l’enzyme bactérienne, Rnt1p lie son substrat via quatre motifs de liaison à l’ARN, ainsi qu’un cinquième motif impliqué dans la reconnaissance de la tétra-boucle. Ce nouveau motif structurel, appelé « G clamp », interagit avec la guanosine conservée, expliquant ainsi la spécificité particulière de Rnt1p pour les tétra- boucles NGNN. Étonnamment, des interactions ont aussi été observées entre le sommet de la tétra-boucle et le domaine N-terminal de Rnt1p, ce dernier étant retrouvé sous la forme d’un dimère. Ces résultats suggèrent que Rnt1p utilise deux « règles moléculaires » pour mesurer la distance entre la tétra-boucle et le site de coupure de l’ARN : la première est dépendante du domaine de liaison à l’ARN double-brin et la seconde du domaine N-terminal. En conclusion, cette étude a permis d’expliquer comment le complexe catalytique des RNase III eucaryotes est assemblé, en plus d’éclaircir les mécanismes qui contrôlent la spécificité et la précision du clivage de l’ARN par Rnt1p.

ARTICLE 2 :

Structure of a Eukaryotic RNase III Postcleavage Complex Reveals a Double- Ruler Mechanism for Substrate Selection

Yu-He Liang1, Mathieu Lavoie2, Marc-Andre Comeau2, Sherif Abou Elela2 and Xinhua Ji1

1Biomolecular Structure Section, Macromolecular Crystallography Laboratory, National Cancer Institute, Frederick, MD 21702, USA. 2RNA Group/Groupe ARN, Département de microbiologie et d'infectiologie, Université de Sherbrooke, Sherbrooke, Québec J1E 4K8, Canada

Received August 14, 2013; Revised December 23, 2013, Accepted February 27, 2014, Published April 3, 2014

Summary

Ribonuclease III (RNase III) enzymes are a family of double-stranded RNA (dsRNA)-specific endoribonucleases required for RNA maturation and gene regulation. Prokaryotic RNase III enzymes have been well characterized, but how eukaryotic RNase IIIs work is less clear. Here, we describe the structure of the Saccharomyces cerevisiae RNase III (Rnt1p) post-cleavage complex and explain why Rnt1p binds to RNA stems capped with an NGNN tetraloop. The structure shows specific interactions between a structural motif located at the end of the Rnt1p dsRNA-binding domain (dsRBD) and the guanine nucleotide in the second position of the loop. Strikingly, structural and biochemical analyses indicate that the dsRBD and N-terminal domains (NTDs) of Rnt1p function as two rulers that measure the distance between the tetraloop and the cleavage site. These findings provide a framework for understanding eukaryotic RNase IIIs.

Introduction

Members of the RNase III family are involved in RNA interference (RNAi) (Hutvagner and Zamore, 2002) and RNA maturation (Court, 1993; Lamontagne et al., 2001; Nicholson, 1999). RNase IIIs are found in all species except archaebacteria (Lykke-Andersen and Garrett, 1997). The family members are identified by their homology to the Escherichia coli RNase III (Robertson et al., 1968) and can be divided into four classes (Lamontagne et al., 2001; Nicholson, 2003) based on protein features (Figure 1A). Typically, class I enzymes possess a single RNase III domain (RIIID) followed by a dsRNA-binding domain (dsRBD); class II is defined by the presence of an N-terminal domain (NTD), a RIIID, and a dsRBD; class III enzymes contain N- terminal P-rich and RS-rich domains followed by two RIIIDs and a dsRBD (Lee et al., 2003); and class IV possesses N-terminal helicase, DUF283, and PAZ domains followed by two RIIIDs and a dsRBD (Zhang et al., 2002). It is the N-terminal extension (NTE) beyond the RIIID that distinguishes eukaryotic RNase IIIs (classes II-IV) from bacterial enzymes (class I).

Dicer produces small interfering RNA (siRNA) and microRNA (miRNA) involved in mediating RNAi. The precursor of siRNA is long dsRNA, whereas that of miRNA is pre-miRNA, a stem-loop structure produced by Drosha from an RNA primary transcript known as pri-RNA. Bacterial RNase III can function as either a processing enzyme (Court, 1993) or a binding protein (Oppenheim et al., 1993). As the only RNase III protein in Saccharomyces cerevisiae (Sc), Rnt1p is involved in the production of small nuclear RNA (snRNA) (Abou Elela and Ares, 1998), pre- rRNA (Elela et al., 1996), and small nucleolar RNA (snoRNA) (Chanfreau et al., 1998). It is also involved in the degradation of several mRNAs (Danin-Kreiselman et al., 2003; Meaux et al., 2011). Deletion of RNT1 perturbs the cell cycle and growth, increases the telomere length, inhibits ribosome production, impairs cell- wall stress response, and induces temperature sensitivity (Catala et al., 2008). While Rnt1p normally cleaves local stem-loop structure, it was shown that the enzyme could direct cleavage in trans by binding to small guide RNA in analogy to 75

the mechanism of RNAi, which is absent in S. cerevisiae (Lamontagne and Abou Elela, 2007).

The structure and mechanism of bacterial RNase III have been well characterized (Court et al., 2013; Gan et al., 2006, 2008; Nicholson, 2011), whereas those of the eukaryotic enzymes remain unclear. The structure and mechanism of bacterial RNase III cannot explain the function of features unique for eukaryotic enzymes like the NTE and the two additional side chains found in the cleavage site (Du et al., 2008; Weinberg et al., 2011). Furthermore, several eukaryotic enzymes developed affinity for substrates with a specific structure and sequence, which cannot be explained by the basic mechanism of dsRNA recognition found in bacterial enzyme. For example, Rnt1p does not cleave dsRNA under physiological conditions, but specifically recognizes short RNA hairpins with conserved NGNN tetraloops (G2 loop; Lebars et al., 2001). Substrate recognition by Rnt1p is mediated by dsRBD, while the cleavage is performed at a fixed distance from the tetraloop by RIIID (Ghazal and Elela, 2006; Lamontagne and Elela, 2001, 2004; Lamontagne et al., 2000, 2003). However, it is not clear how Rnt1p selects these substrates and how the enzyme positions its cleavage site at a fixed distance from the tetraloop. Until now, the exact contribution of the tetraloop, especially the conserved guanine nucleotide, to substrate binding and cleavage is still being debated (Ghazal and Elela, 2006; Hartman et al., 2013; Lavoie and Abou Elela, 2008; Wang et al., 2011; Wu et al., 2004).

Here, we report the crystal structure of Rnt1p in complex with a G2 loop, explain how the catalytic complex of eukaryotic RNase III is assembled, and explain the basis of Rnt1p affinity for short RNAs with specific structure and sequence. Surprisingly, a structural motif near the C terminus of dsRBD specifically recognizes the conserved guanine in the second position of the tetraloop. Intriguingly, the NTD acts as a second ruler that measures, together with the dsRBD, the distance between the tetraloop and the cleavage site. 76

Figure 1. Mode of Substrate Recognition by RNase III. (A) Domain structures are illustrated for Homo sapiens Dicer (HsDicer, UniProtKB Q9UPY3), Giardia intestinalis Dicer (GiDicer, UniProtKB A8BQJ3), Homo sapiens Drosha (HsDrosha, UniProtKB Q9NRR4), Saccharomyces cerevisiae Rnt1p (ScRnt1p, UniProtKB Q02555), Kluyveromyces polysporus Dcr1 (KpDcr1, UniProtKB A7TR32), Aquifex aeolicus RNase III (AaRNase III, UniProtKB O67082), and Bacillus subtilis Mini-III (BsMini-III, UniProtKB O31418). The scale on top indicates the polypeptide lengths. The red box in RIIID represents the RNase III signature sequence. (B) The Rnt1p:RNA complex contains two NTD dimers (blue/red), one RIIID dimer (cyan/orange), two dsRBDs (pale cyan/light orange), and two 34 nt RNAs (gray/ dark gray) in addition to Mg2+ ions (black) and solvent molecules (not shown). (C) Distinct modes of dsRNA recognition by ScRnt1p (left) and AaRNase III (right) are illustrated. The RIIIDs are shown as ribbon diagrams outlined with transparent molecular surfaces, dsRBDs as solid molecular surfaces, and RNAs as tube-and-stick models. The span of the two dsRBDs along dsRNA is indicated with a double-headed arrow. See also Figure S1.

Results

Binding Mode of dsRNA by RNase III

Recombinant Rnt1p was incubated with a 34 nt RNA stem loop (Figure S1A). The resulting mixture yielded crystals of an Rnt1p:RNA complex. However, the SDS-PAGE analysis of the crystal content indicated that the protein was degraded into two fragments (Figure S1B), a 14 kDa fragment containing the NTD (residues 42–151) and a 32 kDa fragment containing the RIIID and dsRBD (∆NTD, residues 197–457). The asymmetric unit contained an intertwined dimer of NTD, a ∆NTD, and a 34 nt RNA (Figure S1C). It has been shown that the dimerization of RIIID is required for dsRNA processing by bacterial RNase III in that side chains from both RIIID subunits are involved in the cleavage of each RNA strand, and the cleavage of both strands creates the characteristic 2 nt 3’ overhang in the product (Gan et al., 2006). Like bacterial RNase III, Rnt1p also functions as a homodimer (Lamontagne et al., 2000). Thus, the biological assembly of Rnt1p was obtained by expanding the asymmetric unit via a crystallographic 2-fold axis. It revealed a ∆NTD dimer in complex with an RNA pseudoduplex, formed by two 34 nt RNAs joined by their 2 nt 3’ overhangs, and two NTD dimers (Figure 1B). Similar to the Aquifex aeolicus RNase III (AaRNase III):RNA complex (Gan et al., 2006), the Sc∆NTD:RNA complex also contained a pseudoduplex RNA formed by two product RNAs (Figure 1C). Unlike bacterial RNase III, however, Rnt1p has the NTD (Figure 1A). The NTD, separated from the ∆NTD due to protein degradation, formed an intertwined dimer recognizing both ends of the ∆NTD:RNA complex (Figure 1B). The extra NTD is due to crystallization condition, but its position does not affect the biochemically predicted positioning of RIIID relative to the cleavage site (vide infra).

The RIIID and dsRBD in both AaRNase III and Rnt1p are connected by a 7- residue flexible linker (Figure S2). The structure of this linker is well defined in the Rnt1p:RNA (this work) and AaRNase III:RNA (Gan et al., 2006) complexes. Comparison between the Sc∆NTD:RNA and AaRNase III:RNA structures shows 78

that while the RIIID dimer of the two enzymes adopt the same position relative to the RNA, the span of the two dsRBDs differs by 66 Ǻ(Figure 1C). This conformation of the enzyme significantly increases the buried surface between protein and RNA, which measures 5,520 Ǻ2 in the case of AaRNse III and 8,372 Ǻ2 in the case of Sc∆NTD. Thus, Rnt1p adopts an RNA-binding mode that permits the recognition of the defining substrate elements without disrupting the classical alignment of dsRNA with the two cleavage sites in the RNase III catalytic complex.

Structure of Eukaryotic RNase III Postcleavage Complex

Unlike bacterial RNase III, the cleavage site of eukaryotic RNase III enzymes contains six conserved side chains (Figure 2A). The Rnt1p:RNA structure shows that the enzyme processes dsRNA by a single RNA cleavage event on each RNA strand to generate products with the 2 nt 3’ overhang (Figure 2B, left). In the postcleavage complex, each cleavage site is composed of the 3’ and 5’ termini of a cleaved RNA strand, two Mg2+ ions, four water molecules, and six catalytic side chains (Figure 2B, right). The distance between the 3’-hydroxyl oxygen and the 5’- phosphate phosphorus is 3.1 Ǻ, which corresponds to the van der Waals distance between these two atoms. This arrangement of the RNA within the cleavage site indicates that the Rnt1p:RNA structure represents a catalytic stage immediately after the hydrolysis of the scissile bond. Statistics for the quality of the structure are listed in Table 1.

Table 1. Data Collection and Structure Refinement Statistics

Data Collection Wavelength (Å) 1.0 Space group C2221 Cell dimensions a, b, c (Å) 158.0, 183.8, 61.3 α, β, γ (˚) 90, 90, 90 Resolution (Å) 40-2.50 (2.65-2.50)a Number of unique reflections 31,010 (4,870) b Rmerge (%) 6.6 (78.4) I/σ 14.45 (1.82) Completeness (%) 98.7 (97.5) Redundancy 4.7 (4.3) Structure Refinement Resolution 40-2.5 (2.63-2.50) c Rwork (%) 21.6 (41.0) d Rfree (%) 23.9 (40.8) Number of Atoms Protein 4,081 RNA 574 Water 98 B factors Protein 81.5 RNA 58.9 Water 62.2 Root-mean-square deviation Bond lengths (Å) 0.003 Bond angles (˚) 0.777 Ramachandran plot Most favored region (%) 92.1 Additional allowed region (%) 6.5 Generously allowed region (%) 0.7 Disallowed region (%) 0.7 Resolution 40-2.5 (2.63-2.50) aValues in parentheses are for the highest resolution shell. b Rmerge = Σ|(I − )|/σ(I), where I is the observed intensity. c Rwork = Σhxl | |Fo| - |Fc| | / Σhxl |Fo|, calculated from working dataset. d Rfree is calculated from 3% of data randomly chosen and not included in refinement.

All six catalytic residues conserved in eukaryotic RNase IIIs (N1, D2, D3, E4, N5, and K6; Figure 2A) were well defined. The side chain of N5 was found to interact with two water molecules and one oxygen of the 5’ phosphate, while the side-chain ε-amino group of K6 interacted with one oxygen of the 5’ phosphate and another from the carboxylic group of D3 (Figure 2B). The interaction of these two side chains with the leaving phosphate group via specific hydrogen bonds indicates that they play important roles in stabilizing the transition state. 80

Superposition of the Rnt1p:RNA and the AaRNase III:RNA (Gan et al., 2008) structures showed identical positioning of the E1, D2, D3, and E4 side chains (Figure 2C). Comparison of the Rnt1p:RNA with three RNA-free eukaryotic RNase IIIs (Macrae et al., 2006; Takeshita et al., 2007; Weinberg et al., 2011) revealed substantial conformational changes of the D2, N5, and K6 side chains upon the formation of cleavage site assembly (Figures 2D–2F), demonstrating that the cleavage site assembly of eukaryotic RNase III is formed and configured for cleavage only when all the components are present.

Figure 2. Cleavage Site Assembly of Eukaryotic RNase III. (A) Structure-based sequence alignment of ScRnt1p (this work), KpDcr1 (Protein Data Bank ID [PDB ID] 3RV0), GiDicer (PDB ID 2FFL), HsDicer 81

(PDB ID 2EB1), and AaRNase III, (PDB ID 2NUG). The RIIIID signature sequence is indicated, and the conserved amino acid residues in the cleavage site are highlighted (in red). The boxed residues, N5 and K6, are unique for eukaryotic enzymes. (B) The dsRNA processing center of Rnt1p has two cleavage sites (left). Stereoview on the right shows the cleavage site assembly. Electron density (annealed omit Fo-Fc map, contoured at 4.5 s) is shown as gray nets for the Mg2+ ions (black spheres) and coordinating water molecules (red spheres). Residues are illustrated as sticks in atomic colors (N in blue, C in green [protein] or orange [RNA], O in red, P in orange, and Mg in black). Mg2+ coordination is indicated by solid lines, hydrogen bonds by dashed lines, and the distance between the 3’-hydroxyl oxygen of Cyt34 and the phosphorous of Cyt1 (3.1 Ǻ) by a double-headed arrow. (C–F) Stereoviews show the superposition of the Rnt1p cleavage site assembly (in green) on that of AaRNase III (C), KpDcr1 (D), GiDicer (E), or HsDicer (F). The AaRNase III, KpDcr1, GiDicer, and HsDicer structures were adopted from the PDB as specified in (A) and shown in magenta. See also Figure S2.

RNA-Binding Motif Unique to Rnt1p

Bacterial RNase III recognizes dsRNA with four RNA-binding motifs (RBMs), RBMs 1 and 2 in the dsRBD and RBMs 3 and 4 in the RIIID (Figure 3A, left). RBMs 1 and 3 form multiple hydrogen bonds with the O2’ hydroxyl groups, while RBMs 2 and 4 each project a loop into the minor groove. These four RBMs are conserved in Rnt1p (Figure 3A, right). Surprisingly, the Rnt1p:RNA structure revealed another RBM that is specialized in the recognition of the conserved guanine in the G2 loop, which we named RBM0 (Figure 3B). The RBM0 is formed by two additional a helices (α3, α4) near the C terminus of dsRBD (Figure 3C). The α3 helix was also seen previously in the structures of truncated dsRBD of Rnt1p in complex with RNA (Wang et al., 2011; Wu et al., 2004), but α4 was not observed before. These two helices transform the conserved αβββα fold of dsRBD (Ramos et al., 2000) into a αβββααα fold. Thus, Rnt1p recognizes its substrate using five RBMs (0, 1, and 2 in dsRBD and 3 and 4 in RIIID), corresponding to residues 445–455, 370– 381, 392–399, 265–269, and 292–312, respectively (Figure S2). 82

Figure 3. RNA-Binding Motif RBM0 Identifies the NGNN Tetraloop. (A) The arrangement of the RBMs of AaRNase III (PDB ID 2EZ6) is shown on the left, and that of ScRnt1p (this work) is shown on the right. Only one subunit of the protein dimer is shown. The RIIID is shown in cyan and dsRBD in pale cyan. The RBMs are in blue and the linker between the two domains in red. Stem-loop RNA is shown as a molecular surface in gray or light gray. (B) RBM0 identifies the AGUC tetraloop. The dsRBD, outlined 83

with a transparent molecular surface, is shown as a ribbon diagram with the RBM0 highlighted in blue. The RNA backbone is shown as a ribbon diagram with the four nucleotides in the tetraloop as sticks colored by atom (N in blue, O in red, C in gray, and P in orange). (C) On the left, dsRBD proteins used in structural analysis (in cyan/blue, this work; in yellow, PDB ID 1T4L; in pink, PDB ID 2LBS) are illustrated with RBMs indicated with dashed boxes. On the right, the three dsRBD structures are superimposed. (D) Interaction map for the binding of RBMs 0 and 1 to the AGUC tetraloop via interaction(s) with the base, ribose, and/or backbone of the RNA. Residues of RBMs 0 and 1 are colored in blue and cyan, respectively. (E) RBM0 is required for substrate cleavage. The Long-G2 or Short-G2 substrate was incubated alone (no enz), with recombinant Rnt1p (Rnt1p), or with enzymes lacking the RBM0 (∆RBM0). The reactions were carried out in multiple-turnover (substrate excess) and physiological-salt (150 mM KCl) conditions. S and P indicate the position of intact substrate and cleavage product, respectively. The asterisk indicates a secondary cleavage product. The sizes of the different RNA fragments are indicated on the left. The substrates are illustrated on top. (F) RBM0 is required for substrate binding. Increasing amounts of Long-G2 or Short-G2 were injected into surface-bound, full-length, or ∆RBM0 enzyme. Shown is the binding curve from surface plasmon resonance. The ratio of the resonance unit change (RU) over the theoretical maximal RU (Rmax) obtained for each binding assay is presented in the form of a graph. Error bars show SD from the means for three replicate experiments. See also Figure S3 and Table S2.

Sequence-Specific Tetraloop Recognition by dsRBD

The Rnt1p:RNA structure showed that the dsRBD recognizes the AGUC tetraloop with RBMs 0 and 1 (Figure 3B). As summarized in Figure 3D, a total of 10 hydrogen bonds are formed between the tetraloop and residues in RBM0 (R445, I448, R450, S453, and V454) and RBM1 (Q373, Y375, S376, G379, and A381), demonstrating that the tetraloop is required for recognition. Four of the ten hydrogen bonds are base specific, including one between Ade15 and RBM1 residue Q373 (Figure S3A) and three between Gua16 and RBM0 residues R445, I448, and S453 (Figure S3B). In contrast, interactions of Uri17 with RBM1 (Figure S3C) and of Cyt18 with RBM0 (Figure S3D) are not base specific. As illustrated in Figure S3E, Gua16 is the most conserved, Ade15 is less conserved, and Uri17 and Cyt18 are not conserved. It has been shown previously that Gua16 is essential for 84

both binding and cleavage of G2 substrates, whereas mutations of the other three nucleotides reduce, but do not block, cleavage (Lamontagne et al., 2003, 2004). Thus, the requirement of an individual nucleotide in the tetraloop for cleavage appears to be dictated by its degree of base-specific interactions.

To evaluate the biochemical significance of RBM0, we prepared a truncated Rnt1p that lacks residues 446–471 (∆RBM0). It was purified (Figures S1D and S1E) and tested for RNA binding and cleavage, using G2 substrates with either long or short RNA duplex (Long-G2 or Short-G2; Figure S1A) under multiple-turnover (RNA excess) and physiological-salt (150 mM KCl) conditions. As shown in Figure 3E, Rnt1p was able to cleave both substrates efficiently, while ∆RBM0 did not, indicating that RBM0 is required for Rnt1p function. To distinguish between defects in catalysis and RNA binding, we examined the impact of deleting the RBM0 on RNA binding in the absence of Mg2+ using surface plasmon resonance. As shown in Figure 3F, full-length Rnt1p efficiently bound to both G2 substrates, whereas the ∆RBM0 did not bind the Short-G2 and bound very poorly to the Long-G2. The weak binding detected with the Long-G2 is consistent with a specific role of RBM0 in recognizing the tetraloop and suggests that the other four RBMs may recognize the RNA duplex, albeit inefficiently. We conclude that the RBM0 of Rnt1p is required for substrate binding and cleavage.

Clamp-Shaped Pocket Tailored for the Selection of Unpaired Guanosines

Inspection of the structure near the tetraloop revealed that the conserved guanosine (Gua16) is snuggly positioned into a clamp-shaped pocket formed by one RBM1 residue (G379) and seven RBM0 residues (R445, A446, A447, I448, P449, R450, and S453), which we call G clamp (Figure 4A). The G clamp structure is tailored to the shape and chemical moieties of the protruding guanosine, and as such it is responsible for the sequence requirement of Rnt1p substrates. Moreover, side-by-side comparison of the Rnt1p:RNAAGUC-capped structure (this work) with the dsRBD:RNAAGAA-capped structure (Wu et al., 2004) shows distinct locations of the 85

guanine base in the tetraloop (Figures S4A and S4B). Therefore, the G clamp not only recognizes the guanosine in the second position of the loop, but also reshapes the loop to better fit the enzyme structure.

To evaluate the effect of the G clamp structure on substrate selection, we substituted the Gua16 base with either adenine or 2-aminopurine, which eliminates its specific interaction with R445 (Figure 4B). The impact on protein binding was then tested using the electrophoretic mobility shift assay (EMSA). Rnt1p efficiently bound the unmodified RNA, but not the RNAs with modified bases in that position. This suggests that other interactions, like those observed with Y46 or those with I448 and S453, are not sufficient for efficient binding to the RNA. To more clearly evaluate the relative contribution of G clamp residues that directly interact with Gua16 to the enzyme function, we replaced residues R445, P449, R450, and S453 individually with alanine or lysine and tested the impact on RNA binding and cleavage (Figures 4C and 4D). The substitution of R445 blocked the cleavage and binding of both G2 substrates. In contrast, the substitution of P449, R450, or S453 did not block, but reduced, the binding and cleavage of the two substrates. As expected, the binding to Short-G2, which limits interactions between the enzyme and the RNA beyond the cleavage site, was more affected by the G clamp mutations than the binding to Long-G2 (Figure 4D). However, the cleavage of Short-G2 was not more affected than the cleavage of Long-G2 by these mutations (Figure 4C), suggesting that enzyme-RNA interactions beyond the cleavage site do not directly contribute to cleavage efficiency. The R445 requirement for cleavage can be explained by its side chain stacking with the Gua16 base, the formation of a hydrogen bond with the Gua16 base, and the formation of a hydrogen bond with the G379 carbonyl that forms a hydrogen bond with the Gua16 O2’-OH (Figure 4A). We conclude that residue R445 is required for substrate binding and cleavage, while the intact structure of the G clamp is necessary for optimal affinity and activity.

Residue Q373 forms a single base-specific hydrogen bond with Ade15, the first nucleotide in the tetraloop (Figure S3A), and the Q373A mutation had similar 86

effects as the P449A, R450A, or S453K mutation that did not block, but reduced, the cleavage of the two G2 substrates (Figures S3F and 4C).

Figure 4. G Clamp Is Tailored for the Recognition of Gua16. (A) Stereoview showing that the conserved Gua16 base is buried in the G 87

clamp formed by eight residues from RMBs 0 and 1. Residues are illustrated as sticks (N in blue, O in red, P in orange, and C in pale cyan [amino acids] or gray [nucleotide]) outlined with electron density map (blue/gray nets, 2Fo-Fc map, contoured at 1.2 s). Dashed lines indicate hydrogen bonds. (B) Guanine-specificity of the G clamp. Rnt1p substrates carrying adenine or 2-aminopurine in the second position of the NGNN loop were tested for binding to Rnt1p using EMSA. The top panels show the structure of the three bases; hydrogen bonds formed between each base and the G clamp are indicated by red and blue arrows. The bottom panels show the binding curves of the three substrates derived from gel shift assays. Error bars show SD from the means for three replicate experiments. (C) Contribution of G clamp residues to RNA cleavage. Four G clamp residues directly interacting with Gua16 of the tetraloop were individually mutated, and the impact on enzyme activities was tested using Long-G2 or Short-G2 as described in Figure 3E. Relative velocities (RV), the cleavage rates obtained with the mutants relative to that obtained by the wild-type, are indicated at the bottom. (D) Impact of G clamp mutations on substrate binding. The binding kinetics of different mutations to Long- G2 or Short-G2 were assessed using surface plasmon resonance, and the ratio of the resonance unit (RU) over the theoretical maximal RU (Rmax) obtained for each binding assay is presented in the form of a graph. Error bars show SD from the means for three replicate experiments. See also Figure S4 and Tables S1 and S2.

Rnt1p Interacts with the RNA Stem Upstream of the Cleavage Sites

The conserved RBMs of Rnt1p (RMBs 1–4, Figure S2) interact with the first five base pairs, the ninth base pair, and the scissile bond downstream of the tetraloop, demonstrating that the structural basis for recognizing dsRNA is conserved between bacterial and yeast enzymes. The Rnt1p:RNA structure shows that RBM1 interacts with base pairs 1–4, the α2 helix of the dsRBD interacts with base pair 5 (Figure 5A), RBMs 2 and 4 interact with base pair 9 (Figure 5B), and RBM3 recognizes the scissile bond (not shown). Therefore, using RBMs 0–2 and the α2 helix, the Rnt1p dsRBD interacts with the tetraloop (Figure 3D) and the RNA stem (Figures 5A and 5B) by forming a total of 18 hydrogen bonds to the RNA, confirming that the role of dsRBD in substrate recognition is conserved in Rnt1p. Out of the 18, 17 are directed at the tetraloop and its neighboring 5 base pairs (Figures 3D and 5C), which explains why Rnt1p can bind to a short RNA hairpin 88

with a 5 bp stem (Lamontagne et al., 2003), a substrate much shorter than that to which a bacterial enzyme can bind (Gan et al., 2006).

To determine the significance of the interaction between RBM2 and the ninth nucleotide, which is located midway between the upper stem-loop and cleavage site, we mutated the interacting amino acid K392 and tested the effect on RNA binding and cleavage. Comparison of the cleavage kinetic parameters of the wild-type and the K392A mutant indicated that the mutation did not affect the binding of Long-G2, but reduced the amount of cleaved RNA by ~35% (Figures 5D and 5E). Consistently, analysis of the kinetic parameters of the Long-G2 cleavage indicated slower catalysis and decreased catalytic efficiency with little effect on the

KM (Table S1). Similarly, the K392A mutation impaired the cleavage of the Short- G2 substrates. However, unlike Long-G2, binding to Short-G2 was significantly inhibited by the K392A mutation (Table S1). This result is consistent with previous studies suggesting that the ninth base pair is important for Rnt1p binding (Lamontagne et al., 2003). We conclude that the interaction of Rnt1p with the middle of the RNA stem is required for optimal binding and cleavage and is particularly critical for forming stable complexes with short RNA substrates.

Figure 5. dsRBD Interacts with the Stem down to the Ninth Base Pair below the Tetraloop. (A) Stereoview showing the interactions between RBM1 and the first four base pairs below the tetraloop and that between the α2 K421 and the fifth base pair. Residues are shown as sticks in atomic colors (N in blue, C in green [protein] or gray [RNA], O in red, and P in orange). Dashed lines indicate hydrogen bonds. A zoomed-out view indicating the position of the RBMs within the Rnt1p:RNA structure is shown on the right. (B) Stereoview showing the interactions between the ninth base pair below the tetraloop and residues K392 from RBM2 and K311 from RBM4. (C) Interaction map between the Rnt1p dsRBD and the first five base pairs below the tetraloop. (D and E) Disruption of RBM2 interaction with the ninth base pair below the tetraloop alters RNA binding and cleavage. Residue K392 was mutated, and the impact of the mutation was tested on substrate cleavage (D) and binding (E) as described in Figures 4C and 4D. Error bars show SD from the means for three replicate experiments. See also Tables S1 and S2.

Interaction between the NTD and RNA Increases Precision of Cleavage Site Selection

Deletion of the Rnt1p NTD impairs cleavage at high-salt conditions, which was suggested to contribute to the stability of the protein-RNA complex (Lamontagne et al., 2000; Lavoie and Abou Elela, 2008). However, the mechanism by which the NTD contributes to substrate selection and catalysis remained unclear. As shown in Figure 6A, the NTD has an all-helical fold that forms an intertwined dimer, as was shown for Kluyveromyces polysporus Dcr1 (KpDcr1; Weinberg et al., 2011). Surprisingly, however, the NTD dimer was found in contact with the AGUC tetraloop via four hydrogen bonds between four amino acid residues (Y46 and H54 from one subunit, K58 and H54 from the other) and three nucleotide residues (Gua14, Ade15, and Gua16). Residue Y46 forms a base- specific hydrogen bond with Gua16. In addition, one of the two H54 side chains stacks with the Ade15 base. This set of NTD-tetraloop interactions has also been observed in two other crystal forms (data not shown), indicating that this arrangement is highly preferred. As Ade15 and Gua16 are conserved in G2 loops (Figure S3E), this finding suggests that the NTD directly contributes to substrate selection. To test this possibility, we examined the impact of deleting the NTD on cleavage when compared to the wild-type and the RBM0 deletion. Cleavage reactions were performed first under low-salt conditions (10 mM KCl) with excess protein to allow the detection of any residual activity. As shown in Figures 6B and 6C, wild-type enzyme cleaved the two G2 substrates almost to completion, whereas both the ∆NTD and ∆RBM0 showed reduced cleavage. Furthermore, the deletion of NTD resulted in the appearance of additional cleavage products, corresponding to the cleavage at noncanonical sites, 5–8 nt below the tetraloop. Interestingly, the Short-G2 showed reduced binding with both the wild-type and ∆NTD, relative to the Long-G2 (Table S1). Consistent results were obtained when the cleavage reactions were carried out under either the low-salt but multiple turnover conditions or the crystallization conditions (data not shown). We conclude 91

that the Rnt1p NTD increases the precision of the cleavage site selectivity by directly interacting with Gua16 and the 5’ end region of the G2 loop.

The structural and functional data suggest that two rulers are dedicated for the mechanism of Rnt1p action: ruler 1 is the RIIID1-dsRBD1 fragment, and ruler 2 is the RIIID2:NTD1/NTD2 complex (Figure 6D). Upon the Rnt1p:RNA complex formation, both rulers are sufficiently ‘‘stiff’’ owing to the protein-protein (RIIID1- RIIID2, RIIID1-dsRBD1, RIIID2-NTD1/NTD2, and dsRBD1-NTD1/NTD2) and protein-RNA interactions. Both rulers interact with the NGNN tetraloop: ruler 1 recognizes the Gua16 base, and ruler 2 secures the Gua16 recognition. Hence, the two RNA strands of the substrate are cleaved accurately 14 and 16 nt, respectively, downstream from the tetraloop.

Figure 6. NTD Dimer Interacts with RNA to Increase Binding Affinity and the Precision of Cleavage Site Selection. (A) The NTD dimer of Rnt1p interacts with three nucleotides within or near the AGUC tetraloop. NTD1 (blue) and NTD2 (red) are shown as ribbons. RNA backbone is shown in orange. Contacting residues are shown as sticks. Dashed lines indicate hydrogen bonds. (B and C) Deletion of the NTD impairs cleavage site selection of G2 substrates. The substrates were labeled at their 5’ (B) or 3’ (C) ends. Wild-type enzyme (Rnt1p), enzyme lacking the RBM0 (∆RBM0), or that lacking the NTD (∆NTD) were incubated with either Long-G2 or Short-G2, and the cleavage products were visualized under single-turnover (enzyme excess) and low-salt (10 mM KCl) conditions. Alternative cleavage sites observed in the absence of the NTD are indicated by #. The cleavage rates are indicated as relative velocities (RV) at the bottom. (D) The double-ruler architecture ensures the cleavage accuracy. The protein is illustrated as a molecular surface and color coded as in (A). The RNA is shown as a cartoon in gray for the tetraloop and in blue and red for the stem. Gua16 is highlighted as a ball-and-stick illustration. See also Figure S5 and Tables S1 and S2. 93

Discussion

The majority of RNase III natural substrates are short RNA molecules with specific features located at a fixed distance from the scissile bonds, such as the 3’ overhang and internal or terminal loops (Nicholson, 2003). In this study, we reveal a mechanism by which a eukaryotic RNase III recognizes a terminal tetraloop sequence in harmony with the formation of the catalytic complex. Recognition of the tetraloop is achieved by an extended dsRBD with two additional a helices, α3 and α4. This αβββααα fold of the dsRBD creates a unique guanosine-specific binding motif that we call the G clamp. Surprisingly, the Rnt1p:RNA structure also revealed specific contacts between the NTD dimer and the 5’ end region of the tetraloop. Deletion of the NTD led to cleavage at alternative site(s), suggesting that this motif increases the precision of the cleavage site selectivity. Together, the work presented here demonstrates a mechanism for RNase III substrate selectivity whereby the dsRBD and the NTD dimer jointly bind the tetraloop and position the RIIID dimer along the stem at a fixed distance from the tetraloop.

Eukaryotic RNase IIIs feature an NTE of variable lengths and domain structures (Figure 1A). Yeast Rnt1p is an excellent model system for studying such NTEs. It contains a single NTD, but the linker between the NTD and RIIID (residues 155–190) is 46 residues long (Figure S2). Protein disorder prediction suggests that this linker is disordered and thus flexible. On one hand, the degradation of Rnt1p into the NTD and ∆NTD fragments is, at least in part, due to the length and flexibility of this linker. On the other hand, it is this long and flexible linker that could allow the NTD dimer to reach and interact with the dsRBD and the AGUC tetraloop. It is not clear, however, which NTD within the NTD dimer is connected to which RIIID in the RIIID dimer. We believe that the intact Rnt1p protein could form the same complex because this complex is in excellent agreement with previous studies, which indicated that the NTD of Rnt1p could interact with itself and the dsRBD (Lamontagne et al., 2000) while contacting the 5’ side of the tetraloop (Lavoie and Abou Elela, 2008). Whether the NTD and dsRBD 94

from the same Rnt1p are involved in the recognition of Gua16 in the AGNN tetraloop remains to be seen.

The Rnt1p:RNA structure explains the capacity of Rnt1p to play a general role in RNA processing and regulation in yeast, despite its high specificity to a given RNA structure. With just one exception, the natural substrates of Rnt1p contain the guanine nucleotide in the second position of the tetraloop (J. Gagnon, M.L., and S.A.E., unpublished data). Therefore, the structural basis for the processing of all but one natural substrates of Rnt1p has been provided by the structure. The one substrate RNA that is not capped with the G2 loop (snR48) contains an adenine nucleotide in the second position (the AAGU tetraloop), for which the binding mode by the enzyme remains to be revealed.

The Presence of NTD Increases the Accuracy and Efficiency of Yeast RNase III

Each eukaryotic RNase III contains an NTE that includes the helicase, DUF283 and PAZ domains in Dicer, the P-rich and RS-rich domains in Drosha, or the NTD in KpDcr1 and Rnt1p (Figure 1A). These domains could specifically interact with RNA and influence cleavage in different ways. It has been suggested that Dicer uses the PAZ domain to recognize the end of dsRNA substrate (Zhang et al., 2004) and uses the helicase domain to recognize the loop/bulge structure of short hairpin RNAs for accurate processing (Gu et al., 2012). The Rnt1p:RNA structure demonstrates that the NTD is a well-defined structural domain that exhibits an all-helical fold and forms an intertwined dimer (Figure 1B), which is similar to that of the KpDcr1 (Weinberg et al., 2011). However, unlike the KpDcr1 NTD dimer, which stacks on the back of its RIIID dimer, the Rnt1p NTD extends away (Figure S5A). On one hand, the length of the linker connecting the NTD to RIIID in KpDcr1 is too short to permit the extended NTD-RIIID arrangement as observed in Rnt1p (Figures S2 and S5A). On the other hand, the Rnt1p NTD dimer may not adopt a KpDcr1-like position because the electrostatic potential between 95

the corresponding surfaces of the NTD and the RIID is not compatible (Figure S5C). The distinct orientations of Rnt1p and KpDcr1 domains reflect the difference between the enzymes’ modes of substrate selection. In the case of Rnt1p, the enzyme extends to recognize the terminal tetraloop with both the NTD and the dsRBD, whereas in the case of KpDcr1, neither the NTD nor the dsRBD are positioned to measure the length from the end of a dsRNA molecule to the cleavage sites (Weinberg et al., 2011).

The Rnt1p:RNA structure indicates a direct role of the NTD dimer in substrate selection. This mode of substrate recognition, in which the NTD assists in selecting the cleavage site at a given distance, is not possible in bacterial RNase III because it lacks an NTE. During evolution, however, eukaryotic RNase IIIs have acquired an NTE to enhance the basic dsRNA substrate specificity conferred by the dsRBD.

How Does Eukaryotic RNase III Process dsRNA?

Despite the wealth of information on the function of eukaryotic RNase III enzymes, the biochemistry of these enzymes remains obscure (Li et al., 2010; MacRae and Doudna, 2007). The Rnt1p:RNA structure represents the catalytic state immediately after cleavage and, as such, provides a unique opportunity to look at how the eukaryotic RIIID dimer cleaves dsRNA. The dimerization of RIIID creates the catalytic valley in which the two cleavage sites form the processing center. Each strand of the bound RNA aligns at a cleavage site, and the two sites are spaced such that the cleavage of the two stands creates the 2-nt 3ʹ overhangs on the product ends (Figure 2). Each cleavage site requires six conserved amino acid side chains, among which the general organization of E1, D2, D3, and E4 is preserved from bacteria to human, whereas the involvement of N5 and K6 is unique for eukaryotes. In the Rnt1p:RNA structure, the N5 and K6 side chains interact with the 5’ phosphate group of the cleaved scissile bond, showing how they are directly involved in the formation of the cleavage site assembly of 96

eukaryotic RNase IIIs. We hypothesize that these two residues participate in the stabilization of the transition state during catalysis.

The Rnt1p dsRBD Exhibits a Mode of Sequence-Specific Selection of dsRNA

The classical mode of dsRNA recognition involves interactions between the dsRBD and the minor groove of dsRNA in order to discriminate between RNA and DNA duplexes. These interactions do not contribute directly to cleavage site selection, but instead increase substrate affinity. As such, E. coli RNase III can accurately cleave dsRNA after the deletion of its dsRBD (Sun et al., 2001); Giardia intestinalis Dicer (GiDicer; Macrae et al., 2006) and Bacillus subtilis Mini-III (BsMini-III; Redko et al., 2008) do not even have a dsRBD (Figure 1A). In contrast, the dsRBD of Rnt1p was shown to be required for the recognition of dsRNA capped with NGNN tetraloop (Nagel and Ares, 2000). However, the basis of this substrate selectivity remained unclear because previous structures using truncated dsRBD did not show that the Gua16 in the tetraloop is specifically recognized by the structural motif we termed the G clamp. The G clamp structure is formed by seven RBM0 and one RBM1 residues, resulting in perfect shape complementarity and base-specific hydrogen bonds (Figure 4). The discovery of this G-specific motif illustrates how the dsRBD may acquire affinity to RNA with specific motif, which opens the door for designing other clamp structures capable of targeting RNAs harboring different tetraloop sequences and broaden the substrate specificity of RNase III.

The 7-residue flexible linker between the RIIID and dsRBD of AaRNase III (Figure S2) allows substantial conformational changes between the two domains, suggesting that RNase III first binds its substrate using one dsRBD, and by free rotation of the RNA-bound dsRBD with respect to the RIIID dimer, subsequently brings the RNA in alignment with the catalytic valley for cleavage (Gan et al., 2005). Similarly, the dsRBD of Rnt1p may also perform the initial recognition of the substrate using RBM0, leading to further binding to the stem through interactions 97

with RBMs 1–4. Indeed, in the absence of Mg2+, Rnt1p formed a stable complex with RNA in a dsRBD-dependent manner (Lavoie and Abou Elela, 2008).

The Double-Ruler Mechanism for Substrate Selection

Soon after the discovery of bacterial RNase III, it was noted that the cleavage products of long dsRNAs by this enzyme are of similar size, giving rise to the idea that the enzyme acts as a molecular ruler measuring the number of nucleotides from one end to its cleavage site (Court, 1993; Nicholson, 1999). However, it was Rnt1p that provided a concrete example of cleavage at a fixed distance from a specific RNA structure, namely the NGNN tetraloop (Chanfreau et al., 2000). This pattern of cleavage is induced by a combination of structural and base-specific interactions with the stem loop as summarized in Figure 7A. The Rnt1p:RNA structure shows that the initial binding and positioning box (IBPB), which consists of the conserved NGNN tetraloop, is recognized by RBMs 0 and 1 as well as the NTD dimer, the binding stability box (BSB) is recognized by RBM1 and the α2 helix of dsRBD, and the cleavage efficiency box (CEB) is recognized by RBM3. In addition to these established RNA epitopes (Lamontagne et al., 2003), the structure identifies the middle box (MB) as a site for the interaction of RBMs 2 and 4 with the ninth base pair below the tetraloop. We conclude that the MB is required for stabilizing the protein:RNA complex, especially when the substrate is too short to allow all the RBMs of both subunits to contact the RNA.

Together, the structural and functional information suggests a mechanism of substrate recognition, the double-ruler mechanism, which uses the NTD dimer as a second ruler to ensure the precision of cleavage (Figure 6D). The additional cleavage sites on the G2 substrates observed in the in vitro cleavage reaction catalyzed by ∆NTD (Figures 6B and 6C) are summarized in Figure 7B. This unusual double-ruler mechanism for substrate selection represents an example of molecular ruler evolution and provides a framework for understanding the mechanism of eukaryotic RNase III action. 98

Figure 7. Double-Ruler Mechanism for Substrate Selection by Rnt1p. (A) Schematic representation of Rnt1p interactions with G2 substrates. Nucleotides are shown as rectangles, while RBMs and NTDs are shown as ellipsoids. C1 and C2 indicate the two canonical cleavage sites. The four boxes of the RNA substrate are outlined with dashed lines. Sites of hydrogen bonds formed between proteins and RNA backbone are shown as shaded rectangles and labeled with residue numbers only, whereas nucleotides forming base-specific hydrogen bonds are labeled by both one-letter code and residue number. Lowercase letters c, f, and s indicate positions of chemical interference (Ghazal and Elela, 2006), hydrogen bond requirement for cleavage (Lavoie and Abou Elela, 2008), and sites where hydrogen bonds were observed with the ribose 2’-OH in the dsRBD:RNA structure (PDB ID 1T4L), respectively. (B) Schematic representation of the ∆NTD interactions with the G2 substrates. C1 and C2 indicate the two canonical cleavage sites by Rnt1p. C* signifies additional cleavage sites produced by the ∆NTD.

Experimental procedures

Crystal Structure Determination

The Rnt1p:RNA complex was made by incubation prior to crystallization. Initial screening for crystallization conditions was carried out with a Hydra II Plus One robot system (Matrix Technologies Corporation). X-ray diffraction data were collected at the SER-CAT 22-ID beamline of the Advanced Photon Source and processed using program XDS (Kabsch, 2010). The structure was solved by molecular replacement (MR) and difference Fourier synthesis. X-ray data and structure refinement statistics are listed in Table 1. The MR search was very difficult due to protein degradation. For details, see Supplemental Experimental Procedures.

In Vitro Cleavage Assays

In vitro cleavage reactions were conducted as described (Lavoie and Abou Elela, 2008). For a brief description, see Supplemental Experimental Procedures.

Electrophoretic Mobility Shift Assays

EMSA assays were performed as described (Lamontagne et al., 2001) with less than 10 fmols of radiolabeled RNA and Rnt1p concentrations ranging from 500 to 4,000 nM. Kd values were calculated using a single binding site nonlinear regression model. The experiment was repeated three times.

Surface Plasmon Resonance

RNA binding over Rnt1p was monitored using the T200 Biacore system (GE Healthcare Life Sciences) in 30 mM HEPES (pH 7.5), 150 mM KCl, and 10 nM EDTA. Different versions of His-tagged Rnt1p (Lamontagne et al., 2001) were immobilized on Ni-NTA chip via the His-tag prior to RNA injection. The surface was washed with 350 mM EDTA and 0.1% SDS between each sample. Steady-state levels of binding were calculated relative to the amount of protein bound to the chip for each RNA concentration to generate resonance unit change (RU) over the 100

theoretical maximal RU (Rmax) values. The experiment was repeated three times.

Kd values were calculated using a single binding site nonlinear regression model and presented in Table S1.

Accession Numbers

The Protein Data Bank (PDB) accession number for the Rnt1p:RNA structure reported in this paper is 4OOG.

Supplemental Information

Supplemental Information includes Supplemental Experimental Procedures, five figures, and two tables and can be found with this article online at http:// dx.doi.org/10.1016/j.molcel.2014.03.006.

Acknowledgments

We thank Catherine Desrosiers for help with protein purification, Jules Gagnon for tetraloop sequence conservation analysis, and Donald Court, George Mackie, and Alexander Wlodawer for discussion. X-ray diffraction data were collected at the SER-CAT 22-ID beamline of the Advanced Photon Source, Argonne National Laboratory. This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research (X.J.) and a grant from the Canadian Institute of Health Research (S.A.E.).

101

References

Abou Elela, S., and Ares, M., Jr. (1998). Depletion of yeast RNase III blocks correct U2 3' end formation and results in polyadenylated but functional U2 snRNA. EMBO J. 17, 3738-3746. Abou Elela, S., Igel, H., and Ares, M., Jr. (1996). RNase III cleaves eukaryotic preribosomal RNA at a U3 snoRNP-dependent site. Cell 85, 115-124. Catala, M., Tremblay, M., Samson, E., Conconi, A., and Abou Elela, S. (2008). Deletion of Rnt1p alters the proportion of open versus closed rRNA gene repeats in yeast. Mol. Cell. Biol. 28, 619-629. Chanfreau, G., Buckle, M., and Jacquier, A. (2000). Recognition of a conserved class of RNA tetraloops by Saccharomyces cerevisiae RNase III. Proc. Natl. Acad. Sci. USA 97, 3142-3147. Chanfreau, G., Legrain, P., and Jacquier, A. (1998). Yeast RNase III as a key processing enzyme in small nucleolar RNAs metabolism. J. Mol. Biol. 284, 975- 988. Court, D. L. (1993). RNA processing and degradation by RNase III, In Control of Messenger RNA Stability, J. G. Belasco, and G. Brawerman, eds. (New York: Academic Press), pp. 71-116. Court, D. L., Gan, J., Liang, Y.-H., Shaw, G. X., Tropea, J. E., Costantino, N., Waugh, D. S., and Ji, X. (2013). RNase III: Genetics and Function; Structure and Mechanism. Annu. Rev. Genet. 47, 405-431. Danin-Kreiselman, M., Lee, C. Y., and Chanfreau, G. (2003). RNAse III-mediated degradation of unspliced pre-mRNAs and lariat introns. Mol. Cell 11, 1279-1289. Du, Z., Lee, J. K., Tjhen, R., Stroud, R. M., and James, T. L. (2008). Structural and biochemical insights into the dicing mechanism of mouse Dicer: A conserved lysine is critical for dsRNA cleavage. Proc. Natl. Acad. Sci. USA 105, 2391-2396. Gan, J., Shaw, G., Tropea, J. E., Waugh, D. S., Court, D. L., and Ji, X. (2008). A stepwise model for double-stranded RNA processing by ribonuclease III. Mol. Microbiol. 67, 143-154. Gan, J., Tropea, J. E., Austin, B. P., Court, D. L., Waugh, D. S., and Ji, X. (2005). Intermediate states of ribonuclease III in complex with double-stranded RNA. Structure (Camb) 13, 1435- 1442. Gan, J., Tropea, J. E., Austin, B. P., Court, D. L., Waugh, D. S., and Ji, X. (2006). Structural insight into the mechanism of double-stranded RNA processing by ribonuclease III. Cell 124, 355-366. 102

Ghazal, G., and Elela, S. A. (2006). Characterization of the reactivity determinants of a novel hairpin substrate of yeast RNase III. J. Mol. Biol. 363, 332-344. Gu, S., Jin, L., Zhang, Y., Huang, Y., Zhang, F., Valdmanis, P. N., and Kay, M. A. (2012). The loop position of shRNAs and pre-miRNAs is critical for the accuracy of dicer processing in vivo. Cell 151, 900-911. Hartman, E., Wang, Z., Zhang, Q., Roy, K., Chanfreau, G., and Feigon, J. (2013). Intrinsic dynamics of an extended hydrophobic core in the S. cerevisiae RNase III dsRBD contributes to recognition of specific RNA binding sites. J. Mol. Biol. 425, 546-562. Hutvagner, G., and Zamore, P. D. (2002). RNAi: nature abhors a double-strand. Curr. Opin. Genet. 12, 225-232. Kabsch, W. (2010). XDS. Acta Crystallogr. D 66, 125-132. Lamontagne, B., and Abou Elela, S. (2007). Short RNA guides cleavage by eukaryotic RNase III. PLoS One 2, e472. Lamontagne, B., and Elela, S. A. (2001). Purification and characterization of Saccharomyces cerevisiae Rnt1p nuclease. Methods Enzymol. 342, 159-167. Lamontagne, B., and Elela, S. A. (2004). Evaluation of the RNA determinants for bacterial and yeast RNase III binding and cleavage. J. Biol. Chem. 279, 2231- 2241. Lamontagne, B., Ghazal, G., Lebars, I., Yoshizawa, S., Fourmy, D., and Elela, S. A. (2003). Sequence dependence of substrate recognition and cleavage by yeast RNase III. J. Mol. Biol. 327, 985-1000. Lamontagne, B., Hannoush, R. N., Damha, M. J., and Abou Elela, S. (2004). Molecular requirements for duplex recognition and cleavage by eukaryotic RNase III: discovery of an RNA-dependent DNA cleavage activity of yeast Rnt1p. J. Mol. Biol. 338, 401-418. Lamontagne, B., Larose, S., Boulanger, J., and Elela, S. A. (2001). The RNase III family: a conserved structure and expanding functions in eukaryotic dsRNA metabolism. Curr. Issues Mol. Biol. 3, 71-78. Lamontagne, B., Tremblay, A., and Abou Elela, S. (2000). The N-terminal domain that distinguishes yeast from bacterial RNase III contains a dimerization signal required for efficient double-stranded RNA cleavage. Mol. Cell. Biol. 20, 1104- 1115. Lavoie, M., and Abou Elela, S. (2008). Yeast ribonuclease III uses a network of multiple hydrogen bonds for RNA binding and cleavage. Biochemistry 47, 8514- 8526. 103

Lebars, I., Lamontagne, B., Yoshizawa, S., Aboul-Elela, S., and Fourmy, D. (2001). Solution structure of conserved AGNN tetraloops: insights into Rnt1p RNA processing. EMBO J. 20, 7250-7258. Lee, Y., Ahn, C., Han, J., Choi, H., Kim, J., Yim, J., Lee, J., Provost, P., Radmark, O., Kim, S., and Kim, V. N. (2003). The nuclear RNase III Drosha initiates microRNA processing. Nature 425, 415-419. Li, W. M., Barnes, T., and Lee, C. H. (2010). Endoribonucleases - enzymes gaining spotlight in mRNA metabolism. FEBS J. 277, 627-641. Lykke-Andersen, J., and Garrett, R. A. (1997). RNA-protein interactions of an archaeal homotetrameric splicing endoribonuclease with an exceptional evolutionary history. EMBO J. 16, 6290-6300. MacRae, I. J., and Doudna, J. A. (2007). Ribonuclease revisited: structural insights into ribonuclease III family enzymes. Curr. Opin. Struct. Biol. 17, 138-145. MacRae, I. J., Zhou, K., Li, F., Repic, A., Brooks, A. N., Cande, W. Z., Adams, P. D., and Doudna, J. A. (2006). Structural basis for double-stranded RNA processing by Dicer. Science 311, 195-198. Meaux, S., Lavoie, M., Gagnon, J., Abou Elela, S., and van Hoof, A. (2011). Reporter mRNAs cleaved by Rnt1p are exported and degraded in the cytoplasm. Nucleic Acids Res. 39, 9357-9367. Nagel, R., and Ares, M., Jr. (2000). Substrate recognition by a eukaryotic RNase III: the double- stranded RNA-binding domain of Rnt1p selectively binds RNA containing a 5'-AGNN- 3' tetraloop. RNA 6, 1142-1156. Nicholson, A. W. (1999). Function, mechanism and regulation of bacterial ribonucleases. FEMS microbiology reviews 23, 371-390. Nicholson, A. W. (2003). The ribonuclease superfamily: forms and functions in RNA maturation, decay, and gene silencing, In RNAi: A Guide to Gene Silencing, G. J. Hannon, ed. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 149-174. Nicholson, A. W. (2011). Ribonuclease III and the role of double-stranded RNA processing in bacterial systems, In Ribonucleases, A. W. Nicholson, ed. (Berline- Heidelberg: Spinger), pp. 269-297. Oppenheim, A. B., Kornitzer, D., Altuvia, S., and Court, D. L. (1993). Posttranscriptional control of the lysogenic pathway in bacteriophage lambda. Prog Nucleic Acid Res Mol Biol. 46, 37-49. Ramos, A., Grunert, S., Adams, J., Micklem, D. R., Proctor, M. R., Freund, S., Bycroft, M., St Johnston, D., and Varani, G. (2000). RNA recognition by a Staufen double-stranded RNA- binding domain. EMBO J. 19, 997-1009. 104

Redko, Y., Bechhofer, D. H., and Condon, C. (2008). Mini-III, an unusual member of the RNase III family of enzymes, catalyses 23S ribosomal RNA maturation in B. subtilis. Mol. Microbiol. 68, 1096-1106. Robertson, H. D., Webster, R. E., and Zinder, N. D. (1968). Purification and properties of ribonuclease III from Escherichia coli. J. Biol. Chem. 243, 82-91. Sun, W., Jun, E., and Nicholson, A. W. (2001). Intrinsic double-stranded-RNA processing activity of Escherichia coli ribonuclease III lacking the dsRNA-binding domain. Biochemistry 40, 14976-14984. Takeshita, D., Zenno, S., Lee, W. C., Nagata, K., Saigo, K., and Tanokura, M. (2007). Homodimeric structure and double-stranded RNA cleavage activity of the C-terminal RNase III domain of human Dicer. J. Mol. Biol. 374, 106-120. Wang, Z., Hartman, E., Roy, K., Chanfreau, G., and Feigon, J. (2011). Structure of a yeast RNase III dsRBD complex with a noncanonical RNA substrate provides new insights into binding specificity of dsRBDs. Structure 19, 999-1010. Weinberg, D. E., Nakanishi, K., Patel, D. J., and Bartel, D. P. (2011). The inside- out mechanism of Dicers from budding yeasts. Cell 146, 262-276. Wu, H., Henras, A., Chanfreau, G., and Feigon, J. (2004). Structural basis for recognition of the AGNN tetraloop RNA fold by the double-stranded RNA-binding domain of Rnt1p RNase III. Proc. Natl. Acad. Sci. USA 101, 8307-8312. Zhang, H., Kolb, F. A., Brondani, V., Billy, E., and Filipowicz, W. (2002). Human Dicer preferentially cleaves dsRNAs at their termini without a requirement for ATP. EMBO J. 21, 5875-5885. Zhang, H., Kolb, F. A., Jaskiewicz, L., Westhof, E., and Filipowicz, W. (2004). Single processing center models for human Dicer and bacterial RNase III. Cell 118, 57-68.

105

Supplemental Information

Structure of a Eukaryotic RNase III Postcleavage Complex Reveals a Double-Ruler Mechanism for Substrate Selection Yu-He Liang, Mathieu Lavoie, Marc-Andre Comeau, Sherif Abou Elela, and Xinhua Ji

Table S1. Kinetic and Binding Parameters of Rnt1p Proteins on the G2 Substrates, Related to Figures 4, 5, and 6.

Long-G2 -1 -1 -1 Protein Kd (µM) KM (µM) Kcat (min ) Kcat/KM (min µM ) Rnt1p 1.47 ± 0.23 0.20 ± 0.08 4.48 ± 0.48 22.58 ± 9.26 S453K 9.06 ± 1.48 0.39 ± 0.18 0.69 ± 0.11 1.80 ± 0.88 R450A 11.39 ± 1.97 2.73 ± 2.05 1.25 ± 0.54 0.46 ± 0.40 P449A 5.77 ± 0.88 0.59 ± 0.20 3.27 ± 0.39 5.51 ± 1.96 R445A > 20 ND ND ND Q373A 2.09 ± 0.36 0.35 ± 0.13 0.95 ± 0.11 2.71 ± 1.06 K392A 1.88 ± 0.30 0.19 ± 0.04 2.96 ± 0.19 15.91 ± 3.95 ∆NTD 3.14 ± 0.53 ND ND ND ∆dsRBD > 20 ND ND ND ∆RBM0 > 20 ND ND ND Short-G2 -1 -1 -1 Protein Kd (µM) KM (µM) Kcat (min ) Kcat/KM (min µM ) Rnt1p 3.24 ± 0.65 0.28 ± 0.08 3.42 ± 0.28 12.22 ± 3.62 S453K > 20 ND ND ND R450A > 20 ND ND ND P449A > 20 ND ND ND R445A > 20 ND ND ND Q373A 4.65 ± 0.93 0.84 ± 0.18 1.65 ± 0.17 1.97 ± 0.47 K392A 18.02 ± 4.22 ND ND ND ∆NTD 5.61 ± 1.09 ND ND ND ∆dsRBD > 20 ND ND ND ∆RBM0 > 20 ND ND ND ND = not determined.

106

Table S2. List of Oligonucleotides Used For PCR-directed Mutagenesis of Rnt1p, Related to Figures 3, 4, 5, and 6

RNT1-361REV-Sal1 5'- CAT ATC AAG TTT GTC GAC TTA CTC CAA GGC AAC -3' RNT1-445-REV-SAL-pQE31 5'- TTC ACT CCT AGG GTC GAC TTA TCT TTG TTT GGC -3' RNT1-K392E-FOR 5'- AAA GAA CCC ACT GCA GTT GAT CCT AAT TCC ATA G -3' RNT1-P449A-FOR 5'- CAA AGA GCT GCC ATT GCT AGG AGT GAA TCT GT -3' RNT1-Q373A-FOR 5'- ATG AAT GCT AAA AGG GCG CTT TAC TCT TTG ATT -3' RNT1-R445A-FOR 5'- TTT TAC GCC AAA CAA GCA GCT GCC ATT CCT AG -3' RNT1-R450A-FOR 5'- AGA GCT GCC ATT CCT GCG AGT GAA TCT GTG TT -3' RNT1-S453K-FOR 5'- ATT CCT AGG AGT GAA AAG GTG TTA AAA GAT CCC -3' RNT1-FORBamH1-pQE31 5'- CTT TTC AAG GAT CCG GGC TCA AAA GTA GC -3' RNT1-FULL-REV-SAL-pQE31 5'- TGG TTG TGT AAA GTC GAC TTA TCA GCT TGT ATC -3'

107

Figure S1. Composition and Quality of RNA and Protein Molecules Used in This Study, Related to Figure 1. (A) Sequence and secondary structure of the RNA oligos used in this study. The AGUC canonical substrates (G2) were designed to include the two canonical cleavage sites of Rnt1p. The 34-nt RNA was used for crystal structure determination, which represents the Rnt1p cleavage products. Arrowheads indicate Rnt1p cleavage sites. The Long and Short-G2 indicate the sequence of Rnt1p substrates R31D and R31L (Lamontagne and Elela, 2004). (B) Top: The Rnt1p protein used for crystallization was purified on 15S column and the resulting fragments separated on SDS-PAGE (Coomassie staining). The positions of Rnt1p and size markers are indicated on the sides. Lanes 1 and 2 were 1-µl loads from two collection tubes and lane 3 was a 10-µl load from one of the two tubes. Bottom: The Rnt1p protein degraded in the crystallization drops into the NTD and ∆NTD (RIIID-dsRBD) fragments as revealed by the SDS-PAGE (Silver staining) of crystals. Additional bands 108

revealed by silver staining on the gel of crystals (Bottom) may represent low levels of contaminations introduced during protein expression, purification, and/or crystallization that cannot be detected by less sensitive Coomassie staining. (C) The content of the crystallographic asymmetric unit. (D) Schematic representation of different Rnt1p variants used in this study. The RBM positions are showed on top and the number of C- terminal residue for each Rnt1p variant is indicated. (E) The different Rnt1p mutants were purified using affinity chromatography and separated using SDS-PAGE. The size markers are indicated on the left of each gel image.

109

Figure S2. Sequence and Structure of RNase III, Related to Figure 2. Structural-based sequence alignment of ScRnt1p (this work), KpDcr1 (PDB entry 3RV0), and AaRNase III (PDBentry 2EZ6). Structurally unobserved regions are indicated in grey. Residues conserved in two and three sequences are indicated in blue and white, respectively. The NTD, RIIID, and dsRBD of Rnt1p are underlined in blue, cyan, and pale cyan, respectively. RBMs are underlined in blue. The four conserved residues in the cleavage site of all RNase IIIs are marked with red triangles and those found only in eukaryotic cleavage sites are indicated with stars. The secondary structure elements found in the current Rnt1p:RNA structure are indicated on top.

110

Figure S3. Structural Details of Rnt1p-Tetraloop Interactions and Impact of the Q373A Mutation on Substrate Cleavage, Related to Figure 3. (A-D) Interactions between Rnt1p and each of the four nucleotides in the AGUC tetraloop. Amino acids from RBMs 0 and 1 are shown as sticks in atomic color scheme (N in blue, O in red, and C in pale cyan). Nucleotides are shown as thicker sticks with carbon atoms in grey. Dashed lines indicate hydrogen bonds. (E) Tetraloop sequence distribution in 36 natural Rnt1p substrates (28 ncRNA, 2 introns, 5 mRNAs, and NPL3- GPI17 intergenic region). (F) The interaction between Q373 and the first nucleotide of the loop enhances cleavage by Rnt1p. Recombinant Rnt1p and the Q373A mutant were assayed for the cleavage of both Long- and Short-G2 substrates (S) under multiple turnover conditions (RNA excess) and physiological salt concentrations (150 mM KCl). The cleavage products (P) were separated on 20% denaturing acrylamide gel. Schemes representing the substrate properties are shown on top of gels and the fragment sizes shown on the left. The cleavage velocities relative to that of the unmodified enzyme (RV) are shown at bottom.

111

Figure S4. Recognition of the AGUC, AGAA, or AAGU Tetraloop by ScRnt1p, Related to Figure 4. (A-C) Side-by-side comparison of the dsRBD-tetraloop interactions determined in this study (Rnt1p:RNAAGUC) with those previously obtained using truncated dsRBDs (dsRBD:RNAAGAA, PDB entry 1T4L; dsRBD:RNAAAGU, PDB entry 2LBS). The dsRBDs are shown as cartoons (helices as spirals, strands as arrows, and loops as tubes) in pale cyan for dsRBD:RNAAGUC (A), yellow for dsRBD:RNAAGAA (B), or pink for dsRBD:RNAAAGU (C). RBMs are highlighted in blue and the dsRBDs outlined with transparent molecular surfaces in white. The Cα position of P449 is indicated with a sphere as a reference point in the three structures for comparison. Tetraloops are shown as stick models colored by atom (N in blue, O in red, P in orange, and C in pale cyan for AGUC, yellow for AGAA, or pink for AAGU), overlapped with the stem-loop RNAs illustrated as tubes. The second nucleotide in the tetraloop (Gua16 or Ade16) is highlighted as spheres. The double-headed arrows indicate distances between the Cα position of P449 and base of the second nucleotide in the tetraloop of stem-loop RNAs.

112

Figure S5. Distinct Positions of the NTD Dimer in Relation to the RIIID Dimer in KpDcr1 and ScRnt1p, Related to Figure 6. (A) Dimeric structures of the NTD-RIIID fragment of KpDcr1 (in orange and light orange, PDB entry 3RV1, on the left) and that of ScRnt1p (in cyan and pale cyan, this work, on the right) are superpositioned on the basis of Cα positions in the NTD dimer (in the middle). The proteins are shown as cartoons (helices as spirals, strands as arrows, and loops as tubes) with the RIIID dimers outlined with transparent molecular surfaces. (B) The complementary landscape and surface electrostatic potential (positive in blue, negative in red) between the RIIID dimer and the NTD dimer of KpDcr1 (PDB entry 3RV1). (C) Neither the landscape nor the surface electrostatic potential for equivalent surfaces between the RIIID dimer and the NTD dimer of ScRnt1p is compatible for KpDcr1-like back packing.

113

Supplemental Experimental Procedures Sample Preparation for Crystallization The full-length, wild-type Rnt1p (Figure S1B) was prepared as described (Lamontagne et al., 2001) with several modifications to the original protocol. Briefly, the E. coli strain M15 containing the expression vector pQE31-Rnt1p was incubated in a BioFlo 415 Fermentor (New Brunswick Scientific Co., Inc, NJ) to an OD600 value of 1.5 in super broth media containing 100 μg/ml ampicillin, and then induced with 1 mM isopropyl β-d-1-thiogalactopyranoside for additional 4 hr at 37°C. The cell lysis was performed using an APV-2000 homogenizer in buffer A [30 mM Tris HCl (pH 8.0), 1 M NaCl, and 10% glycerol] containing 20 mM imidazole and protease inhibitor cocktails (Roche, CA). After centrifugation at 4°C and 15000 rpm for 30 min, the supernatant was loaded onto a nickel-chelating affinity column (HisPrep FF 16/10, GE Healthcare, CA) previously equilibrated with buffer A containing 20 mM imidazole. The impurities were washed out with buffer A containing 70 mM imidazole and the target protein was eluted with Buffer A containing 150 mM imidazole. Further purification of the dimeric form of Rnt1p was carried out by using a Superdex 200 column (HiLoad 26/60 Superdex 200 pg, GE Healthcare, CA) with an elution buffer B [30 mM Tris HCl (pH 7.5) and 0.5 M NaCl]. The final purification was performed with ion-exchange column (Source 15S, GE Healthcare, CA), and the Rnt1p was eluted with buffer C [50 mM phosphate buffer (pH 7.2), 300 mM NaCl]. The purified protein was dialyzed against buffer D [25 mM Tris HCl (pH 7.2), 200 mM NaCl] and concentrated to a concentration of 15 mg/ml. The RNA oligo 5'-P- CAUGUCAUGUCAUGAGUCCAUGGCAUGGCAUGGC-3ʹ (Figure S1A), which was based on the product of natural cleavage of the U5 snRNA 3ʹ end, was purchased from Dharmacon, Inc, CO. The stock solution was prepared by dissolving the RNA powder in a buffer containing 25 mM Tris HCl (pH 7.2) and 100 mM NaCl to a concentration of 4 mM.

114

Crystallization, Data Acquisition, and Crystal Structure Determination and Refinement

The protein-RNA complex was made by incubation prior to crystallization. The mixed solution, containing 0.1 mM Rnt1p, 0.2 mM RNA, 150 mM KCl, and

25 mM MgCl2 in 25 mM Tris HCl (pH 7.2), was incubated at 32°C for 30 min and then cooled on ice. Initial screening for crystallization conditions was carried out with a Hydra II Plus One robot system (Matrix Technologies Corporation). The hits of crystallization conditions obtained from the initial screenings were optimized and scaled up using the sitting-drop vapor diffusion method. Crystals suitable for X-ray diffraction were grown by mixing 3 μl protein-RNA solution and 1 μl reservoir solution [25% w/v PEG1000 in 0.1 M Tris HCl (pH 8.5)] and the 4-μl droplets were equilibrated against 300 µl reservoir solution. The plate- shaped crystals grew to the full size (~0.2 x 0.2 x 0.02 mm3) in two weeks. They were flash-cooled in liquid nitrogen after a short soak in the reservoir solution premixed with 35% (wt/vol) trehalose as a cryoprotectant. X-ray diffraction data were collected at beamline ID-22 of SER-CAT at the APS, Argonne National Laboratory.

The structure was solved by molecular replacement (MR) using phenix.automr of the PHENIX program suite (Adams et al., 2010). The protein degradation made the MR search very difficult because it was hard to estimate the composition and solvent content of the crystal. The structure was finally determined by using an ensemble of monomeric RIIIDs from seven structures [KpDcr1 (PDB entries 3RV0 and 3RV1), AaRNase III (1YYW and 2NUG), Thermotoga maritime RNase III (1O0W), Campylobacter jejuni Rnc (3N3W and 3O2R), and mouse Dicer (3C4T)] and the Rnt1p dsRBD structure (1T4O) as the search models. The NTD dimer and RNA molecules were manually built into the difference density map. The structure was refined with phenix.refine of PHENIX (Adams et al., 2010) and model building and adjustment was carried out with COOT (Emsley and Cowtan, 2004). About 3% of the reflections were randomly selected for cross-validation. Magnesium ions and water molecules were included 115

on the basis of difference electron density (Fo-Fc, above 3σ) and verified with omit maps. The refined structure was validated using the PROCHECK (Laskowski et al., 1993) and WHATIF (Hooft et al., 1996) programs as well as the validation server of PDB (http://validate.rcsb.org/). X-ray diffraction data and structure refinement statistics are listed in Table 1.

Preparation of Rnt1p Mutants for In Vitro Binding and Cleavage Assays

Point mutations in Rnt1p were generated using PCR-based targeted mutagenesis (Good and Nazar, 1992). Truncated versions of Rnt1p were created by single PCR amplification of the desired region within Rnt1p genomic locus. The ∆NTD mutant was previously described (Lamontagne et al., 2000). Purified PCR fragments were inserted between the BamH1/Sal1 sites of the pQE31 vector (Qiagen, Canada) and transformed into BL21 bacteria. Selected clones were sequenced to ensure that no mutations other than the desired ones were present. Proteins were expressed and purified as described (Lamontagne et al., 2001). Oligonucleotides used are listed in Supplementary Table S2.

In vitro Cleavage Assays

Stem-loop RNA substrates derived from the U5 snRNA 3' end were produced with long (Long-G2) or short (Short-G2) stems, 5ʹ -end-labeled with [γ-

32P]ATP using T4 polynucleotide kinase (NEB, Canada), and purified by 20% PAGE. Cleavage reactions were performed by incubating 30 nM Rnt1p with different substrates for 10 min at 30°C in the reaction buffer [30 mM Tris-HCl (pH

7.5), 5 mM spermidine, 0.1 mM DTT, 0.1 mM EDTA, and 10 mM MgCl2] supplemented with either 10 mM KCl (low salt conditions) or 150 mM KCl (high salt conditions). For single turnover reactions, trace amount of radiolabeled RNA (300 cpm/µl) was used. For multiple turnover reactions, 3.2 µM unlabeled RNA was mixed with the trace amount of radiolabeled RNA and added to the reaction. For determination of the Michaelis-Menten constants, reactions were incubated for 1 min in the presence of 50 nM enzyme with RNA concentration ranging 116

between 0.05 and 3.2 µM. Cleavage products were separated on 20% denaturing PAGE, visualized using a Storm 860 imager (Ge Healthcare, Canada), and quantified using QuantityOne software (Biorad). Size markers were generated by alkaline hydrolysis of 5ʹ -end-labeled RNA substrate. KM and Vmax values were calculated using Michaelis-Menten non-linear regression model. All experiments were repeated at least three times. Kinetic parameters are summarized in Table S1. For ∆NTD, in vitro cleavage assay was also conducted with the G2 substrates that were 3ʹ-end-labeled with [γ-32P]pCp using T4 RNA ligase (NEB, Canada), and purified by 20% PAGE.

Supplemental Discussion Hypothetical Model for Recognition of the AAGU Tetraloop by Rnt1p

To date, a total of 43 stem-loop RNA substrates of yeast Rnt1p are known (Jules Gagnon, Mathieu Lavoie, and Sherif Abou Elela, unpublished data). Among these 43 stem-loops, 42 are capped with an NGNN tetraloop (G2-loop), but snR48 exhibits an AAGU tetraloop (A1-loop, Ghazal and Elela, 2006). The snR48 may represent a second type of substrates, although another A1-loop has not been identified. The requirement for the guanosine recognition by the G- clamp raises the question of how Rnt1p binds and cleaves an A1 loop. Previous studies have shown that the AAGU tetraloop (Figure S4C) adopt a backbone conformation that is similar to that of a G2 loop (Figure S4B) upon binding to the dsRBD of Rnt1p (Wang et al., 2011). Accordingly, comparison of our structure (Figure S4A) with the two previously obtained (Figures S4B, S4C) suggests that the presence of the G-clamp may induce a considerable conformational change in the backbone of the RNA. Moreover, the point mutations in the G-clamp has comparable effects in the A1 (Long- and Short-A1, resulted from substitution of the AGUC tetraloop in the G2 substrates with the AAGU tetraloop, Figure S1A) and G2 substrates (data for the two A1 substrates are not shown), suggesting that the G-clamp may be required for the recognition of both substrate types. Thus, we hypothesize that upon binding of the enzyme to the AAGU tetraloop, 117

the RNA may adopt a conformation that allows the guanosine in the third position of the tetraloop to be recognized by the G-clamp. Recognition of the third nucleotide of the tetraloop rather than the second suggests that the resulting structure of the complex has to be somewhat different between G2 and A1 substrates. This model is in agreement with previous biochemical data that suggest that the binding mode of Rnt1p on the RNA and the hydrogen bond interactions are slightly different between A1 and G2 substrates (Ghazal and Elela, 2006). Additional structural information is essential for the characterization of A1-binding mode by Rnt1p.

Supplemental References

Adams, P. D., Afonine, P. V., Bunkoczi, G., Chen, V. B., Davis, I. W., Echols, N., Headd, J. J., Hung, L. W., Kapral, G. J., Grosse-Kunstleve, R. W., et al. (2010). PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D 66, 213-221. Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr. D 60, 2126-2132. Ghazal, G., and Elela, S. A. (2006). Characterization of the reactivity determinants of a novel hairpin substrate of yeast RNase III. J. Mol. Biol. 363, 332-344. Good, L., and Nazar, R. N. (1992). An improved thermal cycle for two-step PCR-based targeted mutagenesis. Nucleic Acids Res. 20, 4934. Hooft, R. W., Vriend, G., Sander, C., and Abola, E. E. (1996). Errors in protein structures. Nature 381, 272. Lamontagne, B., and Elela, S. A. (2004). Evaluation of the RNA determinants for bacterial and yeast RNase III binding and cleavage. J. Biol. Chem. 279, 2231-2241. Lamontagne, B., Larose, S., Boulanger, J., and Elela, S. A. (2001). The RNase III family: a conserved structure and expanding functions in eukaryotic dsRNA metabolism. Curr. Issues Mol. Biol. 3, 71-78.

118

Lamontagne, B., Tremblay, A., and Abou Elela, S. (2000). The N-terminal domain that distinguishes yeast from bacterial RNase III contains a dimerization signal required for efficient double-stranded RNA cleavage. Mol. Cell. Biol. 20, 1104-1115. Laskowski, R. A., MacArthur, M. W., Moss, D. S., and Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283-291. Wang, Z., Hartman, E., Roy, K., Chanfreau, G., and Feigon, J. (2011). Structure of a yeast RNase III dsRBD complex with a noncanonical RNA substrate provides new insights into binding specificity of dsRBDs. Structure 19, 999-1010.

119

CHAPITRE III

Caractérisation des signaux de dégradation par la ribonucléase III Rnt1p à travers le génome de Saccharomyces cerevisiae

AVANT PROPOS

Transcriptome wide annotation of eukaryotic RNase III reactivity and degradation signals

Jules Gagnon*, Mathieu Lavoie*, Mathieu Catala, Françis Malenfant, Sherif Abou Elela

* = co-premier auteurs

Article en révision à PloS Genetics, Septembre 2014

Contribution : J’ai réalisé l’ensemble des expériences en laboratoire (essais enzymatiques, buvardages Northern, qPCRs, extensions d’amorces, etc.). Jules Gagnon a réalisé tout le traitement informatique des données (conservation de séquence et de structure des substrats, prédiction in silico des substrats, traitement des données des puces à ADN et du séquençage à haut débit). Mathieu Catala a réalisé les images de microscopie et l’essai d’oscillations du NADH. J’ai contribué avec lui pour la mesure de l’activité des mitochondries et la croissance des cellules en présence d’azote. J’ai préparé l’ensemble des figures et rédigé les sections « matériel et méthodes » et une version initiale de la discussion. Sherif Abou Elela a rédigé le reste du manuscrit initial auquel j’ai apporté des modifications. J’ai contribué aussi à toutes les révisions subséquentes.

120

RÉSUMÉ

La stabilité de l’ARN joue un rôle crucial dans la régulation de l’expression génique, mais les signaux qui dictent la dégradation conditionnelle et spécifique des ARN sont souvent inconnus. D’un autre côté, la RNase III de Saccharomyces cerevisiae Rnt1p participe à la régulation de l’expression de quelques ARNm en clivant une structure d’ARN similaire à celle utilisée pour la maturation des ARNs non-codants. Afin de mieux comprendre le rôle global de Rnt1p dans la régulation de l’expression génique, nous avons entrepris d’identifier et de caractériser les signaux de coupure de Rnt1p à travers le génome entier de la levure. Les résultats montrent que les motifs de reconnaissance de Rnt1p (les tétra-boucles NGNN) sont très largement répandus à travers le génome de la levure. Toutefois, seulement une minorité des transcrits sont coupés directement par Rnt1p in vitro. Ainsi, la réactivité des substrats semble dépendre du contexte dans lesquels sont placés les signaux de reconnaissance. Étonnamment, la spécificité de Rnt1p ne se limite pas qu’aux tétra-boucles NGNN puisque celui-ci clive aussi quelques transcrits qui présentent des structures non-canoniques telles que des tri-boucles ou des penta-boucles. Par ailleurs, plusieurs des sites de clivage nouvellement identifiés sont retrouvés dans des ARNm de gènes impliqués dans le métabolisme des hydrates de carbone et la respiration mitochondriale. L’impact de Rnt1p sur l’expression de ces gènes varie en fonction des conditions de croissance des cellules. En conclusion, les résultats démontrent que Rnt1p a développé un mécanisme flexible de reconnaissance des substrats qui lui permet de discriminer une grande variété de structures d’ARNs. Ils soulignent aussi l’importance de la dégradation sélective de l’ARN pour la régulation de l’expression génique en réponse aux changements environnementaux.

121

ARTICLE 3 :

Transcriptome wide annotation of eukaryotic RNase III reactivity and degradation signals

Jules Gagnon*, Mathieu Lavoie*, Mathieu Catala, Françis Malenfant, and Sherif Abou Elela

* These authors contributed equally to this work

Département de Microbiologie et d’Infectiologie, Centre of Excellence in RNA Biology Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec, Canada J1E 4K8

Running title: Context dependent regulation of RNA stability

Keywords: RNase III, Transcriptome profiling, RNA degradation, dsRNA, dsRNA

binding proteins, carbohydrate metabolism, respiration 122

Abstract

Detection and validation of the RNA degradation signals controlling transcriptome stability are essential steps for understanding how cells regulate gene expression. Here we present complete genomic and biochemical annotations of the signals required for RNA degradation by the dsRNA specific ribonuclease III (Rnt1p) and examine its impact on transcriptome expression. Rnt1p cleavage signals are randomly distributed in the yeast genome and encompass wide variety of sequence indicating that transcriptome stability is not determined by the recurrence of a fixed cleavage motif. Instead, RNA reactivity is defined by the sequence and structural context in which the cleavage sites are located. Reactive signals are often associated with transiently expressed genes and their impact on RNA expression is linked to growth conditions. Together the data suggest that Rnt1p reactivity is triggered by malleable RNA degradation signals that permit dynamic response to changes in growth conditions.

123

Author Summary

RNA degradation is essential for gene regulation. The amount and timing of protein synthesis is determined at least in part by messenger RNA stability. Although RNA stability is determined by specific structural and sequence motif, the distribution of the degradation signals in eukaryotic genomes remains unclear. In this study, we describe the genomic distribution of the RNA degradation signals required for selective nuclear degradation in yeast. The results indicate that most RNAs in the yeast transcriptome are predisposed for degradation but only few are catalytically active. The catalytic reactivity of messenger RNAs were mostly determined by the overall structural context of the degradation signals. Strikingly, most active RNA degradation signals are found in genes associated with respiration and fermentation. Overall, the findings reported here demonstrate how certain RNA are selected for cleavage and illustrated the importance of this selective RNA degradation for fine tuning gene expression in response to changes in growth condition

Introduction

RNA stability is a critical determinant of gene expression required for the adjustment of RNA abundance in response to changes in growth conditions [1]. Alterations of mRNA stability are associated with many gene expression programs like T cell activation [2], response to osmotic shock [3] and change in carbon source [4]. In addition, selective RNA degradation was shown to play a central role in both cellular and organismal development underlining the importance of this process to the gene expression program [5]. However, despite these profound effects on cell function and growth, the mechanisms by which specific transcripts are selected for degradation remain unclear. RNAs with similar degradation or processing signals often display distinct decay profiles and respond to different cellular cues [6]. Attempts to define the features required for selective RNA 124

degradation are seriously hindered by the limited understanding of the ribonucleases involved in those processes.

In general, RNA turnover and quality control are achieved by exoribonucleases which are mostly controlled by the accessibility of the substrate’s 5’ and 3’ ends [7]. On the other hand, conditional degradation of RNA molecules is often triggered by endoribonucleases that accurately identify specific sequences or structures at a particular time or growth condition [8]. The most studied of these selective endoribonucleases are members of the dsRNA specific ribonuclease III (RNase III) family, which was first discovered in bacteria [9]. These ubiquitous enzymes are defined by their homology to structural elements, which include a nuclease domain (RIIID) that exhibits a conserved divalent metal binding motif, and a double-stranded RNA binding domain (dsRBD) [10]. In bacteria, RNase III regulates the expression of many conditionally expressed genes like those implicated in metal transport [11] and fermentative growth [12]. Similarly, baker’s yeast RNase III (Rnt1p) directly cleaves the mRNA of genes implicated in glucose sensing [13,14], cell cycle and cell wall stress response [15]. In metazoans, the RNase III enzymes Drosha and Dicer are required for the processing of the short non-coding RNA needed for sequence specific RNA degradation [16,17].

The sequence and structural features of natural substrates are hard to identify for most RNase IIIs. Studies of E. coli RNase III suggest that substrate selection is influenced by antideterminant nucleotides (nucleotides that deter cleavage) [18]. On the other hand, eukaryotic RNase IIIs possess more specific mechanisms of substrate selectivity. For example, human Dicer recognizes terminal loops and RNA ends and its substrate specificity is modified in vivo by protein factors like TRBP and PACT [19,20]. Similarly, substrate recognition by Drosha requires a combination of RNA structure and chaperon proteins [8,21]. The most selective enzyme among the members of the RNase III family is found in yeast Saccharomyces cerevisiae. Rnt1p prefers short stem loop structures, capped with either NGNN tetraloop (G2-loop) [10,22,23] or AAGU (A1-loop) structures to long RNA duplexes [22]. Deletion or mutation of these loops block cleavage and 125

reduce binding under physiological conditions [10,24]. This apparently strict substrate specificity suggests that Rnt1p has fewer and more homogeneous targets than other RNase III. However, our knowledge of Rnt1p substrates was deduced from a relatively small number of related substrates (e.g. snoRNAs) [25] that may not reflect the entire spectrum of the enzyme reactivity. Indeed, the broad impact of RNT1 deletion on yeast phenotypic behavior and transcriptome suggests that Rnt1p reactivity is not restricted to non-coding RNA processing [13,15]. This is consistent with the fact that Rnt1p is the only homologue of RNase III proteins in S. cerevisiae.

In this study, we used a combination of genome-wide analysis techniques to outline the overall contribution of Rnt1p to the regulation of gene expression in S. cerevisiae and define the nature of its cleavage signals. Direct cleavage assay of the entire transcriptome permitted unbiased characterization of Rnt1p reactivity and defined the predisposition of all transcripts to selective RNA degradation. The results indicate that although Rnt1p cleavage signals are randomly distributed across the yeast genome, only 10% of the genes are upregulated in vivo in the absence of RNT1 and 5% are directly cleaved by the recombinant enzyme in vitro. Many of the newly identified cleavage sites were found in mRNAs associated with nutritional sensing, carbohydrate metabolism and energy production indicating that yeast RNase III is a key regulator of the cell’s response to growth conditions. Surprisingly, Rnt1p cleavage sites were not restricted to fixed loop sequence and size but extended to different types of structures that include stems terminating with tri- and penta-loops with varying sequences. The variety and frequency of the cleavage signal suggest that Rnt1p has developed a flexible substrate recognition mechanism capable of discriminating between a wide-range of structured RNAs, while avoiding the cleavage of duplex RNA, which is the classical target of other members of the RNase III family. This unusual substrate specificity explains how a single RNase III may regulate the expression of single RNA under specific condition [14] with high precision, while retaining the flexibility needed for transcriptome surveillance [26]. 126

Results

In silico queries of NGNN tetraloop structures predict Rnt1p cleavage motifs independent of the RNA context

There are 55 known substrates of Rnt1p (Figure 1A), the majority of which exhibit a well-defined stem loop structure that features an AGNN tetraloop (G2- loop) (Figure 1B). Therefore, we took advantage of the distinct sequence and structural features of the G2-loop to create an algorithm capable of identifying potential Rnt1p cleavage signals across the entire yeast genome in silico (Figure S1A). This algorithm assigns a score (ranging between 0 and 1) to each predicted structure based on sequence conservation, structural stability and similarity to known Rnt1p targets. As shown in Figure 1C, 80% of the known substrates exhibited scores higher than 0.85. On the other hand, substrates not folding into a G2-loop like snR48 or MATa1 intron and those not forming at least three stable base-pairs downstream of the tetraloop (e.g. snR46 and ARN2-1) were given no score. Substrates generated via long-range interaction or based on non-canonical stems (e.g. ADI1, snR59 and snR190) were scored between 0.68-0.81 (Table S1). Accordingly, we chose 0.85 as a cutoff score to retain the majority of known substrates and reduce the number of false positives. Decreasing the cutoff to 0.8 increased the number of detected known substrates by one, while adding 7071 weak hits. On the other hand, increasing the score to 0.9 resulted in the loss of two known substrates and the removal of 4036 putative hits.

Overall, the algorithm identified 254349 possible loops of which 6321 exhibited a score equal to or higher than 0.85 (Figure 1D and Table S2). To validate the reactivity of the predicted cleavage signals and directly evaluate the validity of the cutoff threshold, we synthesized 24 stem-loop structures spanning the score range between 0.85 and 1 and tested them for cleavage in vitro. As indicated in Figure S1B and Table S3, the enzyme cleaved all but four of the tested stem-loop structures. The 4 non-reactive stem-loops did not share similar scores but instead featured wobble base pairing (G-U) in the first 2 positions 127

downstream of the loop (Figure S1 and Table S4). Based on this result we expect that 83% (with a 95% confidence range of 61 to 95%) of the in silico predicted cleavage sites with scores between 0.85 and 1 are cleaved by Rnt1p in vitro. Therefore, while the algorithm may falsely recognize a group of non-reactive stem- loops due to the inclusion of inhibitory features, like non-canonical base pairing, the majority of the predicted loops appear to be cleavable by Rnt1p in vitro.

Analysis of the genomic distribution of the newly identified G2-loops indicated that 46% reside in protein coding genes (PCGs), 44% opposite to an annotated gene (antisense) and 8.2% in intergenic regions, while only 1.1% were detected in non-coding RNA (Figure 1E and Table S4). This distribution reflects the normal repartition of the yeast genome without preferences to transcript type. On average, potential Rnt1p substrates were distributed in the yeast genome every 4 to 6 kb and displayed similar distribution patterns in other genomes and randomized sequence (Figure S1C). Remarkably, the majority of the predicted cleavage motifs were found in untranscribed regions (e.g. antisense, intergenic regions) confirming that the genomic distribution of the cleavage motifs is not driven by RNA expression (Figure 1E and Table S2). Therefore, the high substrate frequency does not indicate a particularly high demand for RNA degradation in yeast but instead reflects the loose features of Rnt1p substrates.

As expected, examination of the 30 sequences showing the highest score revealed strong enrichment in cleavage signals associated with the processing of pre-snoRNAs (Figure 1F), which constitute the majority of transcripts in the algorithm’s training set. Interestingly, 30% of the top scoring stem-loops were found in mRNAs and 77% of these were located in previously unidentified substrates (Table S5). Cleavage of three of the highest scoring mRNAs was tested in vitro (Figure 1G). All three mRNAs (POM33, SYG1 and HSP60) had a loop score > 0.98 and their expression levels varied between 1.3-8.7 copies per cell [27]. Despite the similarity between cleavage motifs, only SYG1 and HSP60 mRNAs were cleaved by Rnt1p suggesting that sequence outside the stem-loop structure may influence substrate reactivity. Consistently, T7 RNA polymerase 128

transcribed version of the POM33 stem-loop was accurately cleaved by Rnt1p when expressed outside its natural mRNA context (Figure S1D). This confirms the accurate prediction of the stem-loop structure and suggests that the lack of cleavage is due to context dependent changes in the stem-loop structure, stability or accessibility. We conclude that RNA degradation in yeast is not limited by the recurrence of Rnt1p cleavage motifs but depends on the surrounding sequence context that influences its formation and reactivity.

Figure 1. RNA degradation is induced by context dependent activation of randomly distributed cleavage motifs. (A) Types of published Rnt1p substrates. Non-coding RNAs (ncRNAs) include snRNAs, snoRNAs and rRNA. Protein coding genes (PCGs) include ORFs, introns and UTRs (see Table S1). (B) Schematic representation of the common 129

features of Rnt1p G2-loop substrates. Unpaired nucleotides are underlined. Black circles indicate preferential nucleotide pairing. Enriched bases are named while N indicates any nucleotide. The cleavage sites are indicated by arrowheads. IBPB, BSB, MB and CEB indicate initial binding and positioning box, binding stability box, middle box and cleavage efficiency box [10,36], respectively. (C) The score of known substrates was calculated (see Figure S1) and displayed as histogram. (D) G2-loops were scored across the genome and the loop frequency indicated as histogram curve. (E) Pie chart illustrating the types of RNA associated with Rnt1p loops. Antisense and LTRs indicates orphan loops located opposite to an annotated gene or found in long terminal repeat elements, respectively. (F) Pie chart illustrating the types of RNA harboring the top 30 scoring loops (see Table S5). (G) Northern blot analysis of Rnt1p cleavage products. RNA extracted from RNT1 and rnt1∆ cells was incubated with recombinant Rnt1p (rnt1∆ + Rnt1p) and the cleavage products visualized using gene specific probes. ACT1 was used as loading control. The position of the rRNA and products (P) are indicated beside each gel. The relative RNA amount (RMA) was determined using quantitative RT-PCR and indicated below the gels.

Deletion of RNT1 perturbs gene expression regardless of substrate reactivity

Ribonuclease dependent changes in RNA expression were previously used to identify RNA degradation targets [28,29]. Therefore, we compared the transcriptome of RNT1 and rnt1∆ cells using tiling arrays (Figure 2) and confirmed the results using quantitative RT-PCR (Figure S2 and Table S6). The coefficient of determination between the array and quantitative RT-PCR data was 0.746, which is reasonable given the great difference in sensitivity of the two techniques. As expected, the expression of most known Rnt1p substrates increased in rnt1∆ cells by > 2 folds (Figure 2A upper panel and Table S1). A minority of known substrates was overexpressed between 1.2 and 2 folds and only one (snR66) [30], which is also processed by other ribonucleases, was not upregulated. Surprisingly, deletion of RNT1 preferentially increased the expression of weakly expressed sequences (e.g. intergenic regions) by > 1.2 fold while reducing the expression of highly expressed genes (e.g. ribosomal protein genes) (Figure 2A upper panel and Table S7). This unusual perturbation of gene expression was not specific to a chromosome or genomic region and was not observed with other nuclear 130

ribonucleases like Rrp6p [31] (Figure 2A lower panel and Table S8). Together the data suggest that Rnt1p has a broad impact on gene expression that favors the repression of scarcely expressed sequences.

Despite the overall deregulation of the yeast transcriptome, only 498 segments (overlapping 721 genes) were upregulated by more than 2 folds in rnt1∆ cells (Table S9). The majority of the upregulated sequences were in protein coding genes (Figure 2B). In contrast, 22 out of the 30 most upregulated sequences in rnt1∆ cells were associated to snoRNA genes and 1 with the snRNA U2 (Figure 2C and Table S10). The big difference in the expression of snoRNAs resulted mostly from the retention of the externally transcribed spacers (ETS) that are normally processed by Rnt1p (Figure S2D and [25]). In contrast, none of the 7 most overexpressed mRNAs were previously identified as Rnt1p targets. Northern blot analysis of three of those mRNAs confirmed the array predicted overexpression, but only RTC3 was cleaved by Rnt1p (Figure 2D). We conclude that mRNA overexpression in rnt1∆ cells does not necessarily predict the enzyme biochemical reactivity but instead mostly identifies genes that are indirectly affected by the deletion of RNT1. 131

Figure 2. Gene expression is a poor indicator of substrate reactivity. (A) Deletion of RNT1 induces global perturbation of the yeast transcriptome. The levels of gene expression in wild type (WT) cells, rnt1∆ (top) and rrp6∆ (bottom) strains were determined using tiling arrays covering the entire yeast genome and presented in the form of dot plot comparison. Dots corresponding to the expression values of protein- coding genes, non-coding RNA and intergenic regions are shown in black, red and green, respectively. Blue crosses highlight the expression of the known Rnt1p substrates as indicated in Figure 1A. (B) Rnt1p regulates the expression of both coding and non-coding genes. Pie chart illustrating the types of RNAs upregulated by at least two folds upon RNT1 deletion (details in Table S9). Multiple transcripts within the same overexpressed segments were counted individually. PCGs, ncRNAs and LTRs indicate protein-coding genes, non-coding RNAs and long terminal repeats, respectively. (C) Expression of non-coding RNAs (ncRNAs) is highly sensitive to RNT1 deletion. The top 30 upregulated genes in rnt1∆ cells were sorted according to their RNA types (details in Table S10). (D) RNAs upregulated by the deletion of RNT1 resist cleavage in vitro. Northern blot analysis of RNA extracted from wild type (RNT1) and rnt1∆ cells before and after incubation with Rnt1p (rnt1∆+Rnt1p) was performed as described in Figure 1G. Cleavage was visualized using probes complementary to the sequence of HSP12 (left), YGP1 (middle) or RTC3 (right) mRNAs. 132

Genome-wide profiling of biochemical reactivity identifies Rnt1p cleavage targets independent of variation in gene expression

Detection of catalytic activity is the best and most direct way to uncover the substrates of any enzyme. Therefore, we have developed a new method termed “Cut and Chip” that permits direct detection of all the RNAs cleaved by Rnt1p in the yeast transcriptome (Figure 3A and Figure S3). In this new method, the 3’ end cleavage products generated by Rnt1p in vitro are degraded by the 5’-3’ exoribonuclease Xrn1p [32] and the decrease in the RNA level is detected using tiling array (Figure S3A). As shown in Figure 3B, 50% cleavage of Rnt1p substrate (MIG2) [13] was easily detected by the decrease in the array signals downstream of the cleavage site in both Rnt1p and Xrn1p dependent manner. Overall, this approach detected cleavage in 237 RNA transcripts (Table S11), which represent 4% of the yeast genes. Cleavage motifs were detected in 79% of the cleaved RNA (Figure S3C) suggesting that the majority of Rnt1p substrates use G2-loops for cleavage. The majority (71%) of the 55 known substrates were positively identified in this assay. However, we found that detection of cleavage events was dependent on the length of the transcribed sequence downstream of the cleavage site (i.e. length of the 3’ end product) and strength of cleavage (Figure S3D). Therefore, weak cleavage events and those producing 3’ end fragments smaller than 50 nucleotides may not be detected by this technique. Interestingly, cleavage of native RNA by Rnt1p in whole cell extracts produced similar results to that detected by the cleavage in vitro (Table S11). We conclude that the reactivity of the majority of Rnt1p targets (~80%) is not significantly modified through the RNA extraction process or concealed by other protein factors. However, it remains possible that the reactivity of certain RNA is affected by cellular compartmentalization or requires other in vivo events such as active transcription. Nonetheless, the results suggest that Rnt1p cleaves at least four times more transcripts than previously demonstrated.

The majority (83%) of the cleaved segments were found in coding sequence (Figure 3C and Table S11). Interestingly, 9 out of the 12 genes longer than 7.5 kb 133

in the yeast genome were efficiently cleaved by Rnt1p (Table S11). Also, only 8 cleavage events were found in introns and the majority of these (7/8) degrades the intron of mRNAs coding for RNA binding proteins. These introns did not encode for snoRNAs suggesting that cleavage in these pre-mRNAs is not part of a processing pathway. The top 30 genes cleaved by Rnt1p included 18 known substrates, 5 new non-coding RNA substrates (e.g. snR85, snR60, snR81, U3b and TLC1) and 7 new mRNAs, 3 of which originate from genes associated with ribosome biogenesis (HAS1, MDN1 and BFR2) (Figure 3D and Table S12). Primer extension analysis of three newly identified substrates confirmed the capacity of “Cut and Chip” to accurately detect Rnt1p reactivity (Figure 3E). However, primer extension also indicated that “Cut and Chip” does not accurately identify the precise site of cleavage. In most cases, the cleavage segment boundary was associated with several G2-loops and did not always coincide with the position of the cleavage site (Figure 3E). This observation was further validated by primer extension of 3 additional “Cut and Chip” predicted substrates (Figure S3E). Therefore, while “Cut and Chip” is a strong predictor of Rnt1p RNA targets, it does not directly identify the sequence of the cleavage site. 134

Figure 3. Gene expression independent identification of RNA degradation targets. (A) Cut and Chip: a strategy for detecting Rnt1p substrates using tiling array. rnt1∆ RNA was extracted, cleaved with recombinant Rnt1p and the cleavage products degraded using Xrn1p. Differences between the treated and untreated RNA was detected using Affymetrix tiling array and the cleavage fragments identified (Figure S3). (B) Cut and Chip accurately identifies Rnt1p targets. Comparison between 135

the Mig2 mRNA degradation pattern [13] detected by Cut and Chip (left panel) and Northern blot (right panel). ACT1 mRNA was included as negative control [13]. The relative levels of RNA detected by the different probes (black dots) are presented below each ORF. The cleavage sites are shown as stem-loops and the cleaved segments indicated by black line. The Northern blot was performed as described in Figure 1G. (C) Pie chart illustrating the types of RNA cleaved by Rnt1p (Table S11). (D) Pie chart illustrating the type of the top 30 cleaved RNAs (Table S12). (E) Identification of Rnt1p cleavage site using primer extension. Reverse transcription was performed using gene specific radiolabelled primers before (rnt1∆) and after (rnt1∆ + Rnt1p) cleavage. NT, M and P indicate no template control, size markers, and the cleavage product 5’ end, respectively. The smoothed cleavage profile is shown relative to the ORF on top. The predicted G2-loops are indicated as stem loops and the detected cleavage site (CS) identified by the arrows. Probes used for primer extension are shown below each ORF.

Sequencing of Rnt1p cleavage products reveals unexpected diversity of substrates structure

To directly detect Rnt1p cleavage site, we developed a Sequencing Assisted Loop Identification (SALI) technique that permits direct identification of Rnt1p cleavage product (Figure 4 and Figure S4). In this method, the internal cleavage fragment released by Rnt1p is directly sequenced permitting the identification of reactive cleavage signal (Figure 4A). An average of 4.2 million sequencing reads were obtained from both control and cleaved RNA and the 32-38 nucleotides-long reads enriched in the cleaved samples were retained (Figure S4B). As expected, the cleaved RNA sample exhibited a net enrichment in reads ranging between 32 and 38 nucleotides, while most of the reads detected in the control RNA were mapped to abundant small RNAs shorter than 150 nucleotides like tRNAs and snoRNAs. Overall, this technique identified 34 out of 55 known Rnt1p cleavage signals (Table S13). Notably, the boundaries of the enriched reads clusters matched almost perfectly with the position of the cleavage sites reported in the literature (Figure S4C). The missing substrates were either poorly cleaved (e.g. HSL1) [15], produced products longer than 38 nucleotides (e.g. snR51) or were expressed at low levels (e.g. RGT1) [14]. The false positive rate of this technique is 136

estimated to be < 7% based on a list of 30 mRNAs, which showed no cleavage by Northern blot (Table S19). Therefore, SALI is a robust tool for the direct detection of highly reactive Rnt1p cleavage sites.

Overall, SALI detected 243 enriched read clusters corresponding to 203 unique targets (Figure S4D and Tables S13 and S14). The cleavage sites were associated with 131 protein-coding genes, 69 non-coding RNA genes and 3 intergenic regions (Figure 4B). The 30 most enriched cleavage products were found associated with 22 non-coding RNAs and 8 mRNAs (Figure 4C and Table S15). The most enriched sequence mapped to the ETS of a previously uncharacterized cleavage site near the H/ACA snoRNA snR85. Northern blot analysis confirmed the cleavage of the snR85 precursor, which accumulates in rnt1∆ cells (Figure 4D left panel). The 8 highest cleaved mRNAs included only 1 known substrate (MIG2) [13] and 7 new targets (HSP60, TUB1, AXL2, MAP2, HXK1, YTA6 and NAR1) (Table S15). Northern blot analysis of three of these RNAs (HSP60, AXL2 and YTA6) confirmed the cleavage predicted by SALI. In addition, both AXL2 and YTA6 mRNAs accumulated in rnt1∆ cells confirming the capacity of SALI to identify biologically relevant Rnt1p mRNA targets.

Surprisingly, only 44% of the newly identified cleavage products formed the NGNN tetraloop structures deemed essential for Rnt1p reactivity (Figure 4E and Table S13). The rest of the cleavage fragments were either unfolded or formed non-canonical stem loop structures. The newly identified structures included stems capped with either AHNN and BHNN tetraloops or loops with sizes varying between 3 and 6 nucleotides (Figure 4E). To verify the reactivity of these new structures, we generated T7 RNA polymerase transcripts representing the different loop structures and tested them for cleavage in vitro. Structures exhibiting AHNN tetraloops, pentaloops, or triloops were successfully cleaved by Rnt1p while those exhibiting BHNN and hexaloops displayed poor or no reactivity (Figure 4F and Figure S4E). Mutations of the AHNN, pentaloops and triloops indicate that both the structure and sequence of the new loops are required for optimal cleavage (Figure 4F and Figure S4E). Notably, replacement of the established U5 snoRNA G2- 137

tetraloop with the newly identified pentaloop structure of OSH6 did not affect Rnt1p cleavage (Figure S4F). This indicates that pentaloop and G2-loop have comparable reactivity and confirm the newly identified structure as robust Rnt1p substrate. Moreover, Rnt1p cleavage was also observed in the host transcripts of SEC26 and OSH6, further confirming their capacity to be cleaved by Rnt1p (Figure S4G). We conclude that Rnt1p substrate selectivity is not limited to NGNN tetraloop, but extends to a broad range of structured RNAs, which can be distinguished from generic RNA duplexes that are not cleaved by Rnt1p [22].

138

Figure 4. Sequence assisted identification of Rnt1p cleavage signals. (A) Strategy for Sequencing Assisted Loop Identification (SALI). In this strategy, rnt1∆ RNA is cleaved by Rnt1p, RNA fragments < 150 nucleotides enriched and the 32-38 nucleotides internal cleavage fragments identified using next generation sequencing (NGS) (Figure S4). (B) Pie chart illustrating the types of RNAs associated with at least one enriched read cluster (details in Table S14). (C) Distribution of the top 30 enriched read clusters by RNA type (details in Table S15). (D) Cleavage 139

of new substrates was verified using Northern blot as described in Figure 1G. ACT1 mRNA and 5S rRNA were included as negative controls. The relative RNA expression (RMA) was calculated as described in Figure 1G. (E) SALI uncovers new classes of Rnt1p substrates. The cleavages products identified by SALI were folded and the resulting structures shown in the form of pie chart. Unfolded RNAs and stems capped with loops larger than 6 nucleotides were classified as “other”. The loop sequence is described using standard single letter nucleotide code (IUPAC). (F) Validation of the new classes of Rnt1p substrates. A representative member of each loop type (unmodif) as well as mutated loop (MutL) and stem (MutS) versions were transcribed and cleaved by Rnt1p. The cleavage products were identified using 20% denaturating PAGE (see also Figure S4E). Substrate (S) and products (P) are indicated on the right. The size markers are indicated on the left. The observed cleavage sites are indicated by arrowheads.

Comparison and validation of the newly identified RNA degradation targets

The candidate substrates generated by the computational analysis, expression array, Cut and Chip and SALI were compared to evaluate the merit of each method. As indicated in Figure 5A, all four methods were able to detect 65- 80% of all known substrates and, in general, were better at detecting non-coding RNAs. The computational approach identified the highest number (90%) of the known non-coding RNA targets, while Cut and Chip identified the highest number (67%) of the known mRNA targets. In general, the methods based on the quality of the stem-loop (e.g. in silico prediction) and the cleavage efficiency (e.g. SALI) were more successful in identifying highly reactive substrates.

Comparison between the results of the expression array, the Cut and Chip and SALI revealed little overlap between the newly identified RNA targets (Figure 5B). Only 1 mRNA and 29 (31%) non-coding RNA targets were detected by all three methods. The lowest overlap was found between the in vitro cleavage (Cut and Chip and SALI) and the expression-based assays (Figure 5B). Analysis of the stem-loop scores associated with RNA identified by each of the three detection methods indicated that, in general, the RNA identified by the expression array have low loop scores while the highest loops scores were found in RNA identified by the 140

in vitro cleavage assays (Figure 5C). This suggests that the RNA identified by the expression array have less potential to carry a reactive cleavages signal than those found with the in vitro cleavage assays. Indeed, out of the 653 RNAs detected by the expression array, only 36 were cleaved in vitro (Figure 5B), suggesting that the majority of these RNAs are indirectly affected by the deletion of RNT1. On the other hand, a large proportion of the in vitro cleavage targets were not identified by the expression array, likely due to the limited sensitivity and the growth conditions. Indeed, several studies show that Rnt1p can affect the expression of its targets in a condition dependent manner [14,15]. Thus, the newly found targets may not accumulate in absence of RNT1 when cells are grown in optimal conditions. To directly evaluate this hypothesis, we tested the expression of 109 randomly selected in vitro cleavage targets that were not identified by the expression array using quantitative RT-PCR under three different growth conditions (Table S16). As indicated in Figure 5D, 74% of the tested RNAs were upregulated by the deletion of RNT1, suggesting that the majority of the cleavage targets are inhibited by Rnt1p in vivo. However, the upregulation of many targets was detected only under specific growth conditions, confirming the condition dependent repression of gene expression by Rnt1p. Notably, the cleavage product of 57% of the highly expressed in vitro cleavage sites could be detected in vivo upon the deletion of XRN1, which normally degrades Rnt1p cleavage products (Figure 5E) [33,34]. Together these data supports the in vivo reactivity of the newly identified cleavage signals.

141

Figure 5. Evaluation of the different methods used for the detection of Rnt1p cleavage targets. (A) Bar graph indicating the percentage of known substrates (Table S1) detected by each substrate detection assay. (B) Venn diagrams showing the number of Rnt1p cleavage targets identified by the different detection methods in protein-coding genes (top panel) or non-coding RNAs (bottom panel). (C) Box plots showing the distribution of the loop scores associated with the targets identified by each detection method. (D) Bar graph showing the percent of Rnt1p in vitro cleavage targets upregulated after the deletion of RNT1 under different growth conditions. The RNA was extracted from RNT1 and rnt1∆ cells grown in media containing dextrose (2% Dex) or galactose (4% Gal) or cells shifted for 20 minutes in nitrogen supplemented media (20 min N2). The expression levels were examined by quantitative RT-PCR using primers complementary to 109 in vitro cleavage targets not detected by the expression array (see Table S16). Genes showing > 1.2 folds difference in expression between rnt1∆ and RNT1 cells with p-value < 0.01 were considered upregulated. The presented data are the average of 142

three independent experiments. (E) Pie chart showing the percentage of Rnt1p cleavage sites associated with the accumulation of 5’-P cleavage products in xrn1∆ / dcp2∆ cells [34].

Rnt1p cleaves stem-loop structures with different sequence and base pairing requirements

The large number of new substrates identified during this study permits better definition of Rnt1p substrates. Sequence comparison of the G2-loop substrates indicated that the ideal consensus sequence of the G2-loop is AGDU (Figure 6A, left panel), confirming the high conservation of the first two nucleotides and suggests that the 3rd and 4th nucleotides of the loop might be also important for the enzyme reactivity. This finding is supported by earlier work indicating that the enzyme interacts and forms hydrogen bonds with these two nucleotides [35,36]. In addition, comparison of the stem sequence revealed preference in the nucleotide adjacent to the loop, which was previously shown to affect cleavage [10,22,35]. The new model of G2-loop also indicated preference for base pairing in the first 7 nucleotides downstream of the tetraloop consistent with previous biochemical studies indicating the requirement of the first 5 base pairs for cleavage by Rnt1p [37]. Unlike the G2-loops, only a slight tendency to base pairing was detected near the A1- and 5nt-loops (Figure 6A middle and right panels). The small number of candidates and high variability in sequence and structures of these new classes of substrates limited our ability to detect statistically enriched features. However, in general all classes of Rnt1p substrates displayed a tendency to form stable structures with an apparent Gibbs energy below -10.0 Kcal/mol.

Comparison between the Rnt1p G2-loops required for the processing of non-coding RNAs (NCG2-loop) to those required for mRNA (MG2-loop) degradation revealed few differences in sequence and structure. Uracil is preferred in the third position of the NCG2-loops, while guanine is predominant at this position of MG2-loops (Figure 6B). Surprisingly the most marked difference between the two groups of G2-loops was the stem base pairing preferences 143

(Figure 6C). The nucleotides near the cleavage sites (positions 9, 10, 43 and 44 in Figure 6C) and those in the middle stem (positions 15 and 16) were preferentially unpaired in NCG2-loops. The increased mispairing in the more reactive NCG2-loop suggests that unpaired cleavage sites increase reactivity. Indeed, forced pairing of these nucleotides within the cleavage efficiency box decreased the catalytic rate without affecting the substrate affinity [10].

144

Figure 6. Rnt1p cleaves different classes of stem-loop structure with diverse sequence and structural requirements. (A) Revised model of Rnt1p cleavage signals illustrating the common features of reactive RNAs. The substrates were classified based on the loop size and sequence and the three classes exhibiting statistically significant features (p-value < 0.05) were illustrated in the form of stem- loop structures. Underlined nucleotides indicate unpaired positions; black circles indicate paired positions. Grey underline, nucleotide or circle indicates the features which are frequently observed, but not statistically enriched. Arrowheads indicate the cleavage sites. IBPB, BSB and CEB indicate initial binding and positioning box, binding stability box, and cleavage efficiency box [10,36], respectively. The median Gibbs energy (∆G) was also calculated within each group of substrates. (B) The processing and degradation signals display different sequence preferences. The sequence of G2-loop substrates required for protein-coding mRNA degradation (MG2-loop; top panel) was compared to those involved in non-coding RNA processing (NCG2-loop; bottom panel) and the information content of each nucleotide position illustrated in the form of a composite bar graph. The shaded and underlined numbers indicate the position of the tetraloop and cleavage sites, respectively. (C) Rnt1p cleavage sites are preferentially base-paired in protein coding substrates. The percent paired nucleotides at each position of Rnt1p cleavage signals was determined in non-coding RNAs (ncRNA), protein-coding genes (PCGs) and randomly generated sequences (Random seq) and presented in the form of a line graph. Arrows indicate differences between the two groups of cleavage signals.

145

Rnt1p preferentially cleaves mRNA associated with carbohydrate metabolism and respiration

The function of Rnt1p dependent genes was examined using gene ontology [38], MIPS database [39] and literature search (Tables S17 and S18). The results indicated that several genes either upregulated or cleaved by Rnt1p are targets genes associated with mitochondrial respiration and carbohydrate metabolism (Figure 7A). Accordingly, we monitored the effects of variation in carbon sources and oxygen levels on the expression of Rnt1p substrates. As shown in Figures 7B and S5A, 8 substrates were repressed by the enzyme in dextrose, while 3 were repressed in galactose (CDC19, PSK2 and TYE7). Interestingly, three genes which were not affected or downregulated in absence of RNT1 in aerobic condition (CDC19, FBA1 and GPM1), showed clear differences in their expression pattern in response to the nitrogen shift (Figure S5B). Together, these data indicate that the switch from respiration to fermentation modifies the expression of a subset of conditionally expressed genes in a Rnt1p dependent manner.

To evaluate the effect of Rnt1p on respiration, we monitored the levels of mitochondrial membrane electrical gradient (ΔΨm) using the Rhodamine 123 stain [40]. As shown in Figure 7C, deletion of RNT1 reduced staining indicating that the enzyme is required for normal respiration. The change in respiration was not due to changes in the number of mitochondria as indicated by the MitoTracker stain (Figure 7D). Transformation of rnt1∆ cells with a plasmid expressing a wild-type allele of RNT1 (pRNT1) completely restored the Rhodamine staining to its normal levels confirming the direct effect of RNT1 expression on respiration (Figures 7C- E). Consistently, epifluorescence imaging indicated that the morphology of mitochondria was altered in rnt1∆ cells (Figure 7F). Interestingly, despite the perturbation of both fermentation and respiration genes (Table S17 and S18), RNT1 deletion did not block growth in either state, but reduced growth in all carbon sources (Figures S5C and S5D). Therefore, Rnt1p repression of gene expression is not essential for either respiration or fermentation but instead appears to be needed fine tuning gene expression between different metabolic states. To 146

evaluate this possibility, we examined the effect of RNT1 deletion on autonomous oscillation. Autonomous oscillations in the concentrations of glycolytic intermediates like NADH reflect the dynamics of control and regulation of this major metabolic pathway required for both respiration and fermentation [41]. Oscillations of both RNT1 and rnt1∆ strains was induced by the addition of glucose and potassium cyanide and recorded as time traces of NADH fluorescence (Figure 7G). As expected, oscillations were clearly observed in RNT1 cells and lasted for about 15 minutes. In contrast, rnt1∆ cells showed weaker response to glucose induction and oscillations ceased just a few seconds after induction. These results indicate that Rnt1p is required for glycolytic oscillations and the efficient coordination of metabolic flux in yeast.

147

Figure 7. Deletion of RNT1 impairs the expression of in vitro substrates associated with respiration and carbohydrate metabolism. (A) RNA degradation signals accumulate in genes associated with carbohydrate metabolism and energy production. Genes affected by Rnt1p were classified using gene ontology (GO) [70], MIPS database [39] and literature search (Table S17 and S18) and those related to respiration and carbohydrate metabolism shown in the form of a bar graph. (B) The expression of Rnt1p substrates was examined in RNT1 and rnt1∆ cells grown on dextrose (2% Dex), galactose (4% Gal) or after different time following the shift from oxygen to nitrogen supplemented media (N2). The relative expression levels were determined using microarray (Expr. Array) or quantitative RT-PCR and presented in the form of heat-map (see also Figure S5). Genes are grouped according to their response to oxygen depletion. The data shown are an average of three biological replicates. (C) RNT1 and rnt1∆ cells transformed with either the empty vector or pRS316 expressing the wild type RNT1 allele (pRNT1) were stained with Rhodamine 123 (Rho123) and analyzed by flow cytometry. (D) RNT1 and rnt1∆ cells were stained using the Mitotracker stain (MTG) and analyzed by cytometry. (E) The ratio of the Rho123 to MTG signals were calculated 148

for each strain and presented in the form of a bar graph. (F) RNT1 (top) and rnt1∆ (bottom) cells were stained with Mitotracker and the mitochondria (shown in green) visualized using epifluorescence. (G) Rnt1p is required for glycolytic oscillations. Glucose-depleted cell suspensions were supplemented with dextrose (Dex) and potassium cyanide (KCN) to induce oscillations and NADH fluorescence was recorded over time using a temperature-controlled spectrofluorometer.

Discussion

In this study, we used in silico (Figure 1), genetics (Figure 2) and biochemical methods (Figures 3 and 4) to define the characteristics and genomic locations of RNA degradation signals. The results indicate that while potential Rnt1p cleavage motifs are evenly distributed across the yeast genome, only few induce the degradation of the host transcript (Figure 1 and Table S2). The abundance of Rnt1p recognition motifs suggests that RNA degradation is not limited by the de novo evolution of the cleavage motif but instead controlled by the overall structure of RNA transcripts. This is supported by the fact that non-reactive cleavage signals may become reactive in different sequence and structural contexts (Figure S1). Therefore, while Rnt1p cannot cleave generic RNA duplexes like other members of the RNase III family [22], it retains broad substrate specificity by recognizing simple and widely distributed motifs. This flexible substrate specificity may explain how S. cerevisiae maintained its symbiotic relationship with the dsRNA killer virus, which confers selective advantage by producing a toxin that kills uninfected strains [42], without limiting the transcriptome surveillance functions of RNase III.

Expression profiling techniques are extensively used to probe the effects of different ribonucleases on RNA stability and gene expression. These techniques permitted the identification of new classes of unstable transcripts like the Rpr6p- dependant cryptic unstable transcripts (CUTs) [43] and Xrn1p-sensitive unstable transcripts (XUTs) [44]. However, the difficulty in distinguishing between the direct and indirect effects of ribonuclease deletions prevented positive identification of 149

catalytically reactive substrates. Indeed, in this study, the results indicate that most transcripts upregulated after RNT1 deletion resisted cleavage in vitro (Table S19). Biochemical assays like PARE (parallel analysis of RNA ends) were developed to identify the degradation targets of miRNAs [45]. In this study, we used analogous approaches that depend on microarray detection (Cut and Chip) and direct sequencing (SALI) for the identification of Rnt1p cleavage products. Both methods accurately identified the majority of the known Rnt1p substrates (Figure 5A) and the newly identified cleavage events were confirmed using standard cleavage assays. These two methods identified distinct sets of Rnt1p substrates (Table S19 and Figure 5B). Cut and Chip mostly detected long 3’ end cleavage products while SALI detected the cleavage of well-defined and highly expressed cleavage signals (see Figures 5C and S6). For example, SALI detected the highly expressed snR62 cleavage signal, while Cut and Chip detected the long internal structure of snR51. In short, Cut and Chip is more effective in detecting poorly cleaved RNA with long 3’ end cleavage fragment, while SALI is more efficient in directly detecting the site of cleavage of highly expressed and highly reactive substrates. It should also be noted that both methods might fail to detect substrates expressed at very low levels or requiring special factors not present in vitro. Therefore, identification of RNA degradation signals may not be achieved by a single technique but requires a number of complementary approaches that together may provide a true portrait of enzymatic reactivity.

Comparison between the in vitro cleavage assay and expression profiling data indicate that the number of RNA that are both cleaved by Rnt1p and overexpressed after the deletion of RNT1 is very small. Indeed, only 36 out of the 296 mRNA cleaved by Rnt1p in vitro are upregulated upon the deletion of Rnt1p in vivo as predicted by the microarray. This discrepancy reflects the effect of the growth condition tested and the limitation of the expression profiling techniques (Figure S6) that requires arbitrary cutoffs and complicated statistical analysis [46- 50]. Indeed, we clearly show that variations of the growth conditions and the use of quantitative RT-PCR substantially increase the number of the in vitro cleavage 150

targets affected by RNT1 deletion in vivo (Figure 5D and Table S6). Therefore, it appears that the in vitro cleavage assays are better indicators of RNA degradation than the expression array. This is supported by the cleavage of native RNA in cell extracts (Table S11) and the detection of many cleavage products in vivo (Figure 5E). However, we cannot rule out the possibility that the reactivity of certain RNAs might be artificially modified in vitro due to the absence of specific in vivo conditions like active transcription and cellular compartmentalization.

The NGNN tetraloops (G2-loop) are required for the cleavage of most known Rnt1p substrates [51]. For the selection of G2-loops, the enzyme uses a double-ruler mechanism [36]. In this study, we extended the AAGU tetraloop (A1- loop) into the AHNN category and identified two new categories of Rnt1p substrates that feature triloops and pentaloops (Figure 4E). The mechanism by which the enzyme recognizes these three categories of substrates remains unclear. Mutational and biochemical analysis of substrates carrying G2- and A1- loop indicate that Rnt1p use a flexible and interchangeable network of nucleotide interactions to identify its substrates with different structures [22,24]. This flexibility may also explain how the enzyme may cleave different structures like triloops and pentaloops. Alternatively, the enzyme may induce fit the new loops for binding and cleavage [52]. In all cases, the discovery of these new substrates indicates that Rnt1p activity is not restricted to a single stem-loop structure but cover a much larger spectrum of structural motifs.

In nature, yeast cells are in constant flux between respiration and fermentation depending on sugar and oxygen levels [53]. These constant changes in the cell’s metabolic state require a highly responsive and dynamic control of gene expression [54,55]. In high concentration of glucose, yeast cells prefer fermentative metabolism to oxidative pathway regardless of oxygen levels [56]. This reversible respiro-fermentative metabolic state is characterized by induction of genes involved in both glucose transport and glycolysis [57] and repression of the TCA cycle and mitochondrial activity [58]. Interestingly, this combined induction of both transport and glycolysis genes were also observed upon deletion of RNT1. 151

Indeed, glucose sensing (e.g. MIG2, MTH1 and RGT1) [14], glucose transport (e.g. HXT9, HXT11, HXT13 and HXT15), glycolysis (e.g. FBP26, GLK1) and electron transport chain genes (QCR7-9 and CYT1) were all upregulated in rnt1∆ cells (Table S7). Overall, there are now a total of 15 genes in these pathways known to be cleaved by Rnt1p in vitro and their expression is deregulated by its deletion in vivo (Figure 7 and [13,14]). Changing the expression of any one of these genes, like for example MIG2, may explain the rnt1∆ phenotypes, like the perturbation of mitochondrial functions [59], the induction of galactose controlled genes [60] and impaired aerobic metabolism [61]. Since Rnt1p is not essential for growth in either aerobic or anaerobic condition, we conclude that the role of Rnt1p is most likely to fine tune the expression of the genes involved in the respiro-fermentative flux.

Methods

Yeast culture and total RNA extraction The wild type haploid strain RNT1 (MATa, lys2∆0, ura3∆0, his3∆200, leu2∆0) and the haploid rnt1∆ strain (MATa, lys2∆0, ura3∆0, his3∆200, leu2∆0, rnt1::HIS3) were generated by the replacement of one RNT1 allele by HIS3 in the diploid strain (MATa/α lys2∆0/lys2∆0 ura3∆0/ura3∆0 his3∆200/his3∆200 leu2∆0/ leu2∆0), followed by spore dissection as previously described [15]. The rrp6∆ strain (MATa, his3Δ1 leu2Δ0, met15Δ0, ura3Δ0, rrp6Δ::KMX4) was taken from the Yeast knock out collection obtained from Open Biosystems [62]. Yeast cells were grown and manipulated using standard procedures [63] in YEP media supplemented with 2% dextrose at 26°C (the permissive temperature for rnt1∆ strains). In Figure 7B, cells were grown in YEP media supplemented with 2% dextrose or 4% galactose. In the case of the nitrogen shift experiments, assays were performed using cells grown in 600 ml of semisynthetic medium (SSD) containing per liter, 3 g of yeast extract, 10 g of dextrose, 0.8 g of NH4SO4, 1 g of KH2PO4, 0.5 g of NaCl, 0.5 g of

CaCl2•2H2O, 0.3 g of MgSO4, 1.1 µg of FeCl3•6H2O, supplemented with amino acids, adenine and uracil at 40 µg/ml, 0.1% (V/V) Tween 80, 20 µg/ml of ergosterol 152

and 350 ppm of antifoam B emulsion (Sigma-Aldrich, St. Louis, MO) [64] using Multifors (Infors Canada, Anjou, QC, Canada) bioreactors. Cells were grown to 0.3

OD600 in air-supplemented media than the gas supply was shifted to nitrogen. Forty ml samples were collected at different time points and the cells rapidly harvested by filtration. In Figures 7C-E, strains were transformed either with an empty plasmid (pRS316) or expressing the wildtype RNT1 allele (pRNT1) and grown in YC-ura media. Growth rates of RNT1 and rnt1∆ cells grown in presence of different carbon sources was performed and calculated as described [65].

In vitro cleavage of total RNA The cleavage assays were performed as previously described [14] with few modifications. Briefly, 30 µg of total rnt1∆ RNA was incubated with 6 pmol of purified Rnt1p for 20 min at 30°C in 100 µl of reaction buffer [30mM Tris–HCl (pH

7.5), 5mM spermidine, 0.1mM DTT, 0.1mM EDTA (pH7.5), 10mM MgCl2, 150mM KCl]. The reactions were stopped by phenol-chloroform extraction, and the RNA recuperated using salted ethanol precipitation.

Synthesis and cleavage of small RNA hairpins RNA substrates were synthesized using T7 RNA polymerase, radiolabeled and cleaved as described [22]. Briefly, trace amount of radiolabelled substrates (150 cpm/µl) was incubated with 30 nM Rnt1p for 10 min at 30°C in 20 µl reaction buffer [30 mM Tris-HCl (pH 7.5), 5 mM spermidine, 0.1 mM DTT, 0.1 mM EDTA,

10 mM MgCl2 and 10 mM KCl]. Cleavage products were separated on 20% denaturing PAGE and visualized using a Storm 860 imager (GE Healthcare).

RNA detection and analysis Northern blots were performed as described [26] using 15 µg of total RNA and a 1% denaturing agarose gel. The RNA was visualized by autoradiography using randomly labeled probes corresponding to each of the genes examined. 5’- end-labeled oligonucleotide probes were used for detecting snR85 and 5S rRNA. The radiolabeled bands were visualized using a Storm 860 scanner (GE 153

Healthcare) and analyzed using the ImageQuant software (Molecular Dynamics). The primer extension reactions were performed as described [14] using gene specific primers. cDNA synthesis and real-time PCR quantification of relative mRNA expression was performed as described [65] using a Biorad CFX384 Real- Time PCR Detection System. The Ct values of each gene were normalized to the values obtained for the ACT1 mRNA in each samples. The change in gene expression was calculated relative to the values obtained for wild-type RNA. All experiments were performed using at least three independent cultures and the PCR reactions conducted in duplicates. The list of oligonucleotides used to generate Northern blot and primer extension probes, as well as qPCR reactions, can be provided upon request.

In silico prediction of Rnt1p cleavage motifs The method used for the prediction of Rnt1p cleavage signals is adapted from an earlier version used for the prediction of snoRNA cleavage signals [25]. The modified algorithm assigned positional weights based on the level of nucleotides conservation in closely related Saccharomyces species (sensu stricto). The outline of the method is shown in Figure S1A. Nucleotide conservation (Figure 6) was calculated using Fisher test, while base pairing conservation was calculated using chi-squared test p-value < 0.05 and both values were adjusted using Bonferroni correction.

Expression array analysis and data treatment The cDNA preparations and biotin end labeling were performed using total RNA by the University of Wisconsin Gene Expression Center (http://www.biotech.wisc.edu/services/gec). Array hybridization was performed at the Centre for Applied Genomics at University of Toronto (http://www.tcag.ca/index.html) according to the protocols supplied with Affymetrix GeneChip WT Double-Stranded Target Assay (without amplification) and Affymetrix GeneChip S. cerevisiae Tiling 1.0R Array (Affymetrix; PN: 900645). Probes were annotated using the S. cerevisiae S288c reference genome (SGD, 154

http://www.yeastgenome.org, August 10th, 2007) as described [17]. The raw microarray data was treated as described [17] with few modifications using the tilling array R package and in-house scripts. Variations in probe intensity were corrected based on predicted ∆G (Figure S2). The top 5% of the probes with high ∆G was removed and the intensities of the remaining probes adjusted to obtain a null slope and a null average variation. The variation in intensity between RNT1 and rnt1∆ samples was analyzed using Huber segmentation algorithm [66] with consideration of the BIC (Bayesian Information Criterion) optimal number of segment. Variations in expression were defined as the median level of all probes within the region. Neighboring segments with more than 2 fold overexpression and less than 48 nucleotides apart were joined together to form a single segment. Segments with less than 12 uniquely matching probes were considered unreliable and removed from further analyses. Overlapping features (e.g. gene name) were identified using SGD reference genome (http://www.yeastgenome.org, August 10th, 2007).

Identification of Rnt1p cleavage events using tiling arrays (Cut and Chip) Fifty µg of total RNA cleaved with recombinant Rnt1p was incubated with 8 µl Terminator™ 5´-Phosphate-Dependent Exoribonuclease (Xrn1p; Epicentre Biotechnologies, Madison, WI) for 90 min in the supplied buffer. The reactions were stopped by phenol-chloroform extraction, and the RNA collected using salted ethanol precipitation. Preparation of the cDNA, biotin labeling and chip hybridization was performed at the Centre for Applied Genomics at University of Toronto (http://www.tcag.ca/index.html) as described above. Microarray data was analyzed as described for expression array using the variation between treated and untreated samples for segmentation. Segments with less than 12 uniquely matching probes were removed. The median and the MAD (median absolute deviation) of the remaining segments were used to choose an appropriate cutoff (median less 1.96 times the MAD = -0.2425). Neighboring segments with a level below the chosen cutoff were grouped if separated by less than 48 nucleotides and regions smaller than 125 nucleotides were removed. The resulting 237 genomic 155

regions were assigned to annotated features in the SGD genome of August 10th, 2007 (http://www.yeastgenome.org).

Isolation and parallel sequencing of Rnt1p cleavage products (SALI) Hundred µg of RNA cleaved with Rnt1p were purified using the mirVana™ miRNA Isolation Kit (Ambion, Life Technologies, Burlington, ON) to isolate RNA shorter than ~150 nucleotides. The enrichment of short RNAs was confirmed using the Agilent 2100 Bioanalyzer Small RNA kit (Agilent technologies, Santa Clara, CA). The cDNA libraries were generated from 500 ng of size selected RNA using the Ion Total RNA-Seq Kit v2. The IonTorrent sequencing data were generated using Ion 318™ Chip Kit and acquired using Ion PGM System and Torrent Suite 2.2 software. 5' adapter sequences were trimmed using cutadapt 1.2rc2 [67]. Sequences shorter than 16 nucleotides were removed and the remaining reads aligned to S. cerevisiae reference genome sequence R64-1-1 using subread 1.3.5p4 [68]. Sequences with multiple matching positions were removed and reads ranging between 32 and 38 nucleotides were considered for further analysis. Read clusters consisting of 14 or more identical reads found in the cleaved and not the control samples were considered enriched. Enriched clusters of identical reads with over 50% overlap were merged and the resulting clusters were assigned to the transcripts with corresponding sequence. The RNA secondary structure for the longest merged cluster was predicted using Vienna RNA tools version 1.8.5. Wobble base pairs and non-canonical A-C base pairs were permitted in the predicted structures.

Detection of Rnt1p cleavage in whole cell extracts

Exponentially growing rnt1∆ cells were harvested and washed twice in AGK buffer (10 mM HEPES pH 8.0, 1.5 mM MgCl2, 200 mM KCl, 10% Glycerol) containing protease inhibitors. Cell pellet was resuspended in 1 volume of the same buffer and the slurry was quickly frozen in liquid nitrogen. About 12 grams of the frozen cell suspension was lysed in a 6870 Freezer Mill (SPEX SamplePrep). 156

Grinded powder was then thawed on ice and spun at 20 000 g for 30 minutes. The supernatant (S20 fraction) was collected and stored in aliquots at -80°C. Cleavage reactions were performed as described above except that total RNA was replaced by S20 extract (the amount was determined based on the measured RNA content in the extract). Detection of the cleavage products was performed as described for the Cut and Chip method.

Accession numbers and raw data access

Raw and processed data presented in this study was deposited in the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/), under accession number GSE57450.

Detection of Rnt1p cleavage product in vivo

The terminal 5’-phosphates of the 3’ cleavage products identified by SALI and those near the stem-loops predicted by Cut and Chip were compared to those detected in the xrn1∆ / dcp2∆ cells using global 5’ RACE (5’ RACE data was obtained from [34]). The RACE detected 5’-phosphates found within the first 5 nucleotides of the 3’ end of the cleavage products were considered a match and used for the generation of Figure 5E.

Cytometric assessment of respiratory competency Yeast mitochondrial membrane potential was evaluated as previously described with few modifications [69]. Briefly, 4 x 106 cells obtained from an exponentially growing culture were harvested and washed 3 times in PBS solution. Cells were stained with 35 g / ml Rhodamine 123 for 10 minutes at 26°C or with 500 nM MitoTracker green FM for 50 minutes at the same temperature. Stained cells were washed 2 times with PBS and incubated for 15 minutes at 26°C in PBS. 157

The resulting cells were further stained with 0.1g / ml propidium iodide in PBS and analyzed using a Fortessa cytometer (BD Biosciences, Mississauga, ON, Canada) equipped with a 50 mW solid state 488 nm laser. The emitted fluorescence of the Rhodamine123 and the MitoTracker green FM were detected at 530 ± 15 nm, while the propidium iodide detected at 610 ± 10 nm. Forward and side scatter signals were used to exclude debris and cell clumps. Dead cells were identified with propidium iodide staining and excluded from the analysis. For each sample, a minimum of 8 000 positive events by sample were acquired. Fluorescence intensity distribution profiles were traced using CyflogicTM software (CyFlo Ltd, Finland) and raw data were exported and an analyzed as previously described [40].

Microscopy Cells grown to a density of 107 cells / ml in YEPD were stained with 500 nM MitoTracker green FM for 30 minutes at 26°C in growth medium. Stained cells were washed two times in PBS and the mitochondria were visualized using 100 X / 1.46 oil objective with an excitation filter of 470 ± 20 nm and an emission filter of 540 ± 25 nm attached to the Zeiss Axio Observer microscope. Stacks were acquired at 200 nm intervals and deconvoluted using Zeiss Zen iterative algorithm. Maximum intensity projections of the deconvoluted images are shown.

Measurements of NADH oscillations NADH oscillations were measured as previously described [41]. Briefly, 100 ml of YC media buffered at pH 5 with 100 mM potassium phthalate and supplemented with 1% dextrose were inoculated to an OD600 of 0.2 using fresh saturated pre-cultures grown in the same media. Wild type and rnt1∆ cells were grown 16-18 or 36-40 hours, respectively, at 26°C to deplete the sugar in the media. The resulting cells were washed twice with 10 ml of 50 mM potassium phosphate buffer pH 6.8 and finally suspended to 10% wet weight in the same buffer. Suspended cells were incubated 3 hours at 26°C before measuring the NAD / NADH fluorescence in a PTI spectrofluorometer using 3 ml of cell suspension at 30°C in a 4.5 ml PMMA cuvette. Cells were agitated for 5 minutes in 158

the spectrofluorometer before data acquisition. Two readings per seconds were acquired with excitation at 366 nm and emission at 450 nm during 2 minutes for baseline recording before inducing oscillations with 30 mM glucose and 5 mM KCN.

Calculation of gene ontologies enrichment Enriched gene ontologies were detected by standard hypergeometric tests using the GOstats R package (version 2.18.0) [70] and annotation packages version 2.5.0. A Bonferroni corrected p-value of 0.05 was used to select significantly enriched terms. Background gene set contained the top 95% highly expressed mRNAs in rnt1∆ strain.

Acknowledgments

We would like to thank members of the Abou Elela lab, Michelle Scott, Xinhua Ji and Daniel Lafontaine for critical reading of the manuscript. The Ion Torrent sequencing was performed in the Université de Sherbrooke RNomics Platform. 159

References

1. Rabani M, Levin JZ, Fan L, Adiconis X, Raychowdhury R, et al. (2011) Metabolic labeling of RNA uncovers principles of RNA production and degradation dynamics in mammalian cells. Nat Biotechnol 29: 436-442.

2. Cheadle C, Fan J, Cho-Chung YS, Werner T, Ray J, et al. (2005) Control of gene expression during T cell activation: alternate regulation of mRNA transcription and mRNA stability. BMC Genomics 6: 75.

3. Romero-Santacreu L, Moreno J, Perez-Ortin JE, Alepuz P (2009) Specific and global regulation of mRNA stability during osmotic stress in Saccharomyces cerevisiae. RNA 15: 1110-1120.

4. Cereghino GP, Scheffler IE (1996) Genetic analysis of glucose regulation in saccharomyces cerevisiae: control of transcription versus mRNA turnover. EMBO J 15: 363-374.

5. Rodda SJ, Kavanagh SJ, Rathjen J, Rathjen PD (2002) Embryonic stem cell differentiation and the analysis of mammalian development. Int J Dev Biol 46: 449-458.

6. Castilla-Llorente V, Nicastro G, Ramos A (2013) Terminal loop-mediated regulation of miRNA biogenesis: selectivity and mechanisms. Biochem Soc Trans 41: 861-865.

7. Parker R (2012) RNA degradation in Saccharomyces cerevisae. Genetics 191: 671-702.

8. Chakravarthy S, Sternberg SH, Kellenberger CA, Doudna JA (2010) Substrate-specific kinetics of Dicer-catalyzed RNA processing. J Mol Biol 404: 392-402.

9. Nicholson AW (2014) Ribonuclease III mechanisms of double-stranded RNA cleavage. Wiley Interdiscip Rev RNA 5: 31-48.

10. Lamontagne B, Ghazal G, Lebars I, Yoshizawa S, Fourmy D, et al. (2003) Sequence dependence of substrate recognition and cleavage by yeast RNase III. J Mol Biol 327: 985-1000.

11. Lim B, Sim SH, Sim M, Kim K, Jeon CO, et al. (2012) RNase III controls the degradation of corA mRNA in Escherichia coli. J Bacteriol 194: 2214-2220.

12. Membrillo-Hernandez J, Lin EC (1999) Regulation of expression of the adhE gene, encoding ethanol oxidoreductase in Escherichia coli: transcription from a downstream promoter and regulation by fnr and RpoS. J Bacteriol 181: 7571-7579. 160

13. Ge D, Lamontagne B, Abou Elela S (2005) RNase III-Mediated Silencing of a Glucose-Dependent Repressor in Yeast. Curr Biol 15: 140-145.

14. Lavoie M, Ge D, Abou Elela S (2012) Regulation of conditional gene expression by coupled transcription repression and RNA degradation. Nucleic Acids Res 40: 871-883.

15. Catala M, Aksouh L, Abou Elela S (2012) RNA-dependent regulation of the cell wall stress response. Nucleic Acids Res 40: 7507-7517.

16. Lund E, Dahlberg JE (2006) Substrate selectivity of exportin 5 and Dicer in the biogenesis of microRNAs. Cold Spring Harb Symp Quant Biol 71: 59-66.

17. David L, Huber W, Granovskaia M, Toedling J, Palm CJ, et al. (2006) A high- resolution map of transcription in the yeast genome. Proc Natl Acad Sci U S A 103: 5320-5325.

18. Zhang K, Nicholson AW (1997) Regulation of ribonuclease III processing by double-helical sequence antideterminants. Proc Natl Acad Sci U S A 94: 13437-13441.

19. Lee HY, Zhou K, Smith AM, Noland CL, Doudna JA (2013) Differential roles of human Dicer-binding proteins TRBP and PACT in small RNA processing. Nucleic Acids Res 41: 6568-6576.

20. Gu S, Jin L, Zhang Y, Huang Y, Zhang F, et al. (2012) The loop position of shRNAs and pre-miRNAs is critical for the accuracy of dicer processing in vivo. Cell 151: 900-911.

21. Ma H, Wu Y, Choi JG, Wu H (2013) Lower and upper stem-single-stranded RNA junctions together determine the Drosha cleavage site. Proc Natl Acad Sci U S A 110: 20687-20692.

22. Lamontagne B, Abou Elela S (2004) Evaluation of the RNA determinants for bacterial and yeast RNase III binding and cleavage. J Biol Chem 279: 2231- 2241.

23. Lebars I, Lamontagne B, Yoshizawa S, Aboul-Elela S, Fourmy D (2001) Solution structure of conserved AGNN tetraloops: insights into Rnt1p RNA processing. EMBO J 20: 7250-7258.

24. Ghazal G, Elela SA (2006) Characterization of the reactivity determinants of a novel hairpin substrate of yeast RNase III. J Mol Biol 363: 332-344.

25. Ghazal G, Ge D, Gervais-Bird J, Gagnon J, Abou Elela S (2005) Genome- wide prediction and analysis of yeast RNase III-dependent snoRNA processing signals. Mol Cell Biol 25: 2981-2994. 161

26. Ghazal G, Gagnon J, Jacques PE, Landry JR, Robert F, et al. (2009) Yeast RNase III triggers polyadenylation-independent transcription termination. Mol Cell 36: 99-109.

27. Greenbaum D, Jansen R, Gerstein M (2002) Analysis of mRNA expression and protein abundance data: an approach for the comparison of the enrichment of features in the cellular population of proteins and transcripts. Bioinformatics 18: 585-596.

28. Lee A, Henras AK, Chanfreau G (2005) Multiple RNA surveillance pathways limit aberrant expression of iron uptake mRNAs and prevent iron toxicity in S. cerevisiae. Mol Cell 19: 39-51.

29. Danin-Kreiselman M, Lee CY, Chanfreau G (2003) RNAse III-mediated degradation of unspliced pre-mRNAs and lariat introns. Mol Cell 11: 1279- 1289.

30. Lee CY, Lee A, Chanfreau G (2003) The roles of endonucleolytic cleavage and exonucleolytic digestion in the 5'-end processing of S. cerevisiae box C/D snoRNAs. RNA 9: 1362-1370.

31. Burkard KT, Butler JS (2000) A nuclear 3'-5' exonuclease involved in mRNA degradation interacts with Poly(A) polymerase and the hnRNA protein Npl3p. Mol Cell Biol 20: 604-616.

32. Pellegrini O, Mathy N, Condon C, Benard L (2008) In vitro assays of 5' to 3'- exoribonuclease activity. Methods Enzymol 448: 167-183.

33. Meaux S, Lavoie M, Gagnon J, Abou Elela S, van Hoof A (2011) Reporter mRNAs cleaved by Rnt1p are exported and degraded in the cytoplasm. Nucleic Acids Res 39: 9357-9367.

34. Harigaya Y, Parker R (2012) Global analysis of mRNA decay intermediates in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 109: 11764-11769.

35. Lavoie M, Abou Elela S (2008) Yeast ribonuclease III uses a network of multiple hydrogen bonds for RNA binding and cleavage. Biochemistry 47: 8514-8526.

36. Liang YH, Lavoie M, Comeau MA, Abou Elela S, Ji X (2014) Structure of a Eukaryotic RNase III Postcleavage Complex Reveals a Double-Ruler Mechanism for Substrate Selection. Mol Cell.

37. Lamontagne B, Abou Elela S (2007) Short RNA guides cleavage by eukaryotic RNase III. PLoS One 2: e472. 162

38. Osier MV, Zhao H, Cheung KH (2004) Handling multiple testing while interpreting microarrays with the Gene Ontology Database. BMC Bioinformatics 5: 124.

39. Mewes HW, Frishman D, Guldener U, Mannhaupt G, Mayer K, et al. (2002) MIPS: a database for genomes and protein sequences. Nucleic Acids Res 30: 31-34.

40. Ludovico P, Sansonetty F, Corte-Real M (2001) Assessment of mitochondrial membrane potential in yeast cell populations by flow cytometry. Microbiology 147: 3335-3343.

41. Williamson T, Adiamah D, Schwartz JM, Stateva L (2012) Exploring the genetic control of glycolytic oscillations in Saccharomyces cerevisiae. BMC Syst Biol 6: 108.

42. Drinnenberg IA, Fink GR, Bartel DP (2011) Compatibility with killer explains the rise of RNAi-deficient fungi. Science 333: 1592.

43. Wyers F, Rougemaille M, Badis G, Rousselle JC, Dufour ME, et al. (2005) Cryptic pol II transcripts are degraded by a nuclear quality control pathway involving a new poly(A) polymerase. Cell 121: 725-737.

44. van Dijk EL, Chen CL, d'Aubenton-Carafa Y, Gourvennec S, Kwapisz M, et al. (2011) XUTs are a class of Xrn1-sensitive antisense regulatory non-coding RNA in yeast. Nature 475: 114-117.

45. Folkes L, Moxon S, Woolfenden HC, Stocks MB, Szittya G, et al. (2012) PAREsnip: a tool for rapid genome-wide discovery of small RNA/target interactions evidenced through degradome sequencing. Nucleic Acids Res 40: e103.

46. Larsson O, Nadon R (2008) Gene expression - time to change point of view? Biotechnol Genet Eng Rev 25: 77-92.

47. Murie C, Nadon R (2008) A correction for estimating error when using the Local Pooled Error Statistical Test. Bioinformatics 24: 1735-1736.

48. Murie C, Woody O, Lee AY, Nadon R (2009) Comparison of small n statistical tests of differential expression applied to microarrays. BMC Bioinformatics 10: 45.

49. Nadon R, Shoemaker J (2002) Statistical issues with microarrays: processing and analysis. Trends Genet 18: 265-271.

50. Loven J, Orlando DA, Sigova AA, Lin CY, Rahl PB, et al. (2012) Revisiting global gene expression analysis. Cell 151: 476-482. 163

51. Chanfreau G, Buckle M, Jacquier A (2000) Recognition of a conserved class of RNA tetraloops by Saccharomyces cerevisiae RNase III. Proc Natl Acad Sci U S A 97: 3142-3147.

52. Wang Z, Hartman E, Roy K, Chanfreau G, Feigon J (2011) Structure of a yeast RNase III dsRBD complex with a noncanonical RNA substrate provides new insights into binding specificity of dsRBDs. Structure 19: 999-1010.

53. Huberts DH, Niebel B, Heinemann M (2012) A flux-sensing mechanism could regulate the switch between respiration and fermentation. FEMS Yeast Res 12: 118-128.

54. Kim IS, Kim YS, Kim H, Jin I, Yoon HS (2013) Saccharomyces cerevisiae KNU5377 stress response during high-temperature ethanol fermentation. Mol Cell 35: 210-218.

55. Horak J (2013) Regulations of sugar transporters: insights from yeast. Curr Genet 59: 1-31.

56. Raab AM, Hlavacek V, Bolotina N, Lang C (2011) Shifting the fermentative/oxidative balance in Saccharomyces cerevisiae by transcriptional deregulation of Snf1 via overexpression of the upstream activating kinase Sak1p. Appl Environ Microbiol 77: 1981-1989.

57. Johnston M (1999) Feasting, fasting and fermenting. Glucose sensing in yeast and other cells. Trends Genet 15: 29-33.

58. Heyland J, Fu J, Blank LM (2009) Correlation between TCA cycle flux and glucose uptake rate during respiro-fermentative growth of Saccharomyces cerevisiae. Microbiology 155: 3827-3837.

59. Fernandez-Cid A, Riera A, Herrero P, Moreno F (2012) Glucose levels regulate the nucleo-mitochondrial distribution of Mig2. Mitochondrion 12: 370- 380.

60. Lim MK, Siew WL, Zhao J, Tay YC, Ang E, et al. (2011) Galactose induction of the GAL1 gene requires conditional degradation of the Mig2 repressor. Biochem J 435: 641-649.

61. Cao H, Yue M, Li S, Bai X, Zhao X, et al. (2011) The impact of MIG1 and/or MIG2 disruption on aerobic metabolism of succinate dehydrogenase negative Saccharomyces cerevisiae. Appl Microbiol Biotechnol 89: 733-738.

62. Winzeler EA, Shoemaker DD, Astromoff A, Liang H, Anderson K, et al. (1999) Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285: 901-906. 164

63. Guthrie C, Fink GR (1991) Guide to Yeast Genetics and Molecular Biology. San Diego, CA: Academic Press.

64. Burke PV, Kwast KE, Everts F, Poyton RO (1998) A fermentor system for regulating oxygen at low concentrations in cultures of Saccharomyces cerevisiae. Appl Environ Microbiol 64: 1040-1044.

65. Parenteau J, Durand M, Morin G, Gagnon J, Lucier JF, et al. (2011) Introns within ribosomal protein genes regulate the production and function of yeast ribosomes. Cell 147: 320-331.

66. Huber W, Toedling J, Steinmetz LM (2006) Transcript mapping with high- density oligonucleotide tiling arrays. Bioinformatics 22: 1963-1970.

67. Martin M (2011) Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnetjournal 17: 10-12.

68. Liao Y, Smyth GK, Shi W (2013) The Subread aligner: fast, accurate and scalable read mapping by seed-and-vote. Nucleic Acids Res 41: e108.

69. Volejnikova A, Hlouskova J, Sigler K, Pichova A (2013) Vital mitochondrial functions show profound changes during yeast culture ageing. FEMS Yeast Res 13: 7-15.

70. Falcon S, Gentleman R (2007) Using GOstats to test gene lists for GO term association. Bioinformatics 23: 257-258.

165

Supplemental Figures

Figure S1 (related to Figure 1). In silico prediction of Rnt1p cleavage signals. (A) Pipeline for the identification of Rnt1p substrates in silico. The relative weight of each parameter used for determining the final score is shown between parentheses. (B) In silico algorithm identifies Rnt1p reactive stem-loop structures. 24 stem-loops identified in silico spanning scores between 0.85 and 1.00 were T7-transcribed and tested for Rnt1p cleavage in vitro (see also Table S3). The pie chart shows the proportion of targets for which Rnt1p cleavage was observed. (C) RNA degradation is not limited by the evolution of Rnt1p cleavage signals. Rnt1p cleavage signals are not restricted to genomes expressing the enzyme (S. cerevisiae) but extend to genomes expressing enzymes with alternative substrate specificity (S. pombe). Strikingly, the cleavage signals are frequently found in random sequence where di-nucleotide frequency and GC content are comparable to S. cerevisiae genome. The table indicates the total number of signals detected in each genome as well as the average number of loops per kilobase. (D) RNA degradation is determined by the context surrounding the cleavage signal. The predicted stem-loop 166

structure, which failed to induce the cleavage of POM33 mRNA (Figure 1G), was independently produced by T7 RNA polymerase and tested for cleavage by Rnt1p as described in Figure 4F. The position of the substrate (S) and cleavage products (P) is indicated on the right. Size markers (M) are shown to the left of the gel. The predicted hairpin structure is shown on the left and the position of the detected cleavage is indicated by an arrow. The asterisk indicates cleavage at non-canonical sites.

167

Figure S2 (related to Figure 2). Detection and normalization of gene expression profiles. (A) Pipeline for the identification of RNA segments overexpressed in rnt1Δ cells. Data were obtained from Affymetrix yeast genomic tiling arrays hybridized to cDNA generated from wild type and rnt1Δ total RNA. (B) Examples of GC contents adjustments and signal normalization. Signals from the tiling array (i) were adjusted according to the GC content (ii) of each probe and normalized relative to the median signals (iii). The data shown represent the signals obtained from probes hybridizing to the 250 nucleotides before and after MIG2 cleavage site (MIG2 SNR). (C) Validation of the microarray data using quantitative PCR. The expression of 40 genes representing different levels of expression upregulation in rnt1Δ cells was examined using quantitative RT-PCR and presented in the form of a dot plot relative to the array predicted expression levels (see also Table S5). The correlation value between the two methods is shown on top. (D) Rnt1p is required for the removal of snoRNA external transcribed spacers (ETSs). The line graphs illustrate the 168

log2 change in snR48, snR50 and snR69 ETS levels after RNT1 deletion. The nucleotide positions are shown relative to the snoRNA mature 5’ end. A schematic of each gene is shown on top and the position of known Rnt1p cleavage site is indicated as hairpins. 169

Figure S3 (related to Figure 3). Detection and validation of Rnt1p degradation targets. (A) Pipeline for the identification of RNA segments cleaved by Rnt1p in vitro. Data were obtained using Affymetrix yeast genomic tiling array (see Figure. 3A). (B) rnt1∆ RNA incubated in the absence or presence of Rnt1p was separated on agarose gel either directly or after treatment with Xrn1p. The RNA fragments were visualized using probes against two known RNA substrates (MIG2 and GPI17). ACT1 and 5S rRNA were included as negative controls. The uncapped 25S rRNA was used as indicator of Xrn1p activity. Schemes of the mature transcript cleavage products and probe positions (dashed line) are 170

respectively shown on the right. (C) Cleavage segments were separated based on the presence or the absence of G2-loops near the cleavage site and the distribution shown in the form of a pie chart. (D) Examples of RNAs in which reactivity was rated below the Cut and Chip detection cut off (false negatives). Segments were considered to be cleaved by Rnt1p only if their expression profile decreased below -0.2425 or more in order to reduce the number of false positives (see Methods section). However, a few genes like ATP16 and PAM1 showed weak (below cutoff), but distinguishable cleavage patterns upon manual observation. The line graphs show the degradation profile generated from the tiling array data as described in Figure 3E. (E) Additional examples of Cut and Chip predicted substrates validated by primer extension as described in Figure 3E.

171

Figure S4 (related to Figure 4). Detection of Rnt1p cleavage sites using SALI. (A) Pipeline for the detection of internal cleavage segments. (B) Characteristics of the sequencing reads obtained before and after treatment with Rnt1p. (C) SALI accurately detects the position of Rnt1p cleavage site. The sequence obtained by SALI (in bold) of 5 well- established cleavage sites was determined and aligned relative to the above corresponding RNA sequence. Primary and secondary cleavage sites are indicated by large and small arrows, respectively. The tetraloop sequence is underlined. The number of reads corresponding to the sequence shown in bold is indicated on the right. (D) Examples of the read 172

coverage of Rnt1p cleavage signals. The read distribution near Rnt1p cleavage signals is illustrated in the form of a line graph. The nucleotide numbers are indicated relative to the start codon. No corresponding reads were detected in the untreated samples. (E) Additional examples of cleavage sites featuring 4, 5 and 6 nucleotides loops were synthesized using T7 RNA polymerase and tested for cleavage as described in Figure 3F. (F) The G2-loop of the established U5 substrate was replaced by the newly identified 5-nt loop sequence found in OSH6 mRNA and tested for cleavage as described in Figure 4F. The position of the mutations is shown by open and shaded boxes. (G) Northern blot analysis of substrates identified in Figure 4F. Cleavage reactions and Northern blot analysis was performed as described in Figure 1G.

173

Figure S5 (related to Figure 7). Effect of changes in growth conditions on the expression of Rnt1p substrates. (A) Effect of carbon source on the expression of Rnt1p and its substrates. The expression levels were determined using quantitative RT-PCR on RNA extracted from RNT1 (black) and rnt1Δ (grey) cells grown in 4% galactose (Gal) or 2% dextrose (Dex). The data were normalized relative to expression levels of RNT1 cells grown in Dex and shown in the form of a bar graph. (B) Kinetics of gene expression of Rnt1p substrates after oxygen depletion. The expression levels of the different mRNA were determined using quantitative RT-PCR after shift to growth under nitrogen (N2). The expression levels detected in rnt1Δ cells are shown relative to the levels 174

detected in RNT1 strain before N2 shift (t=0 min). (C) The growth rates (in hours) were determined for RNT1 and rnt1Δ cells grown in presence of different carbon sources. The values were calculated from three independent cultures. (D) The growth rates of RNT1 (black) and rnt1Δ (white) cells calculated above were plotted relative to the growth of respective strains in media containing dextrose as carbon source.

Figure S6 (related to Figures 1, 2, 3 and 4). Table comparing the advantages and disadvantages of each of the different methods used to determine Rnt1p cleavage targets.

175

Supplemental Tables

** Étant donné le nombre important de tableaux supplémentaires, ceux-ci n’ont pas été inclus dans le présent document. Ils seront toutefois disponibles pour consultation dans le fichier « Lavoie_Mathieu_PhD_2015_donnees.xls » qui accompagne cette thèse. **

Table S1 (related to Figure 1): List of published Rnt1p substrates considered in this study.

Table S2 (related to Figure 1): Predicted G2-loop cleavage motifs with score above 0.70.

Table S3 (related to Figure 1): In vitro cleavage of T7-transcribed tetraloops.

Table S4 (related to Figure 1): List of transcripts containing at least one predicted G2-loop with scores higher than 0.85.

Table S5 (related to Figure 1): Top 30 G2-loops.

Table S6 (related to Figure 2): Comparison between the tiling arrays and quantitative RT-PCR detected changes in gene expression upon RNT1 deletion

(RNT1 / rnt1∆ expression (log2)).

Table S7 (related to Figure 2): SGD annotated genes overexpressed in rnt1∆ strain.

Table S8 (related to Figure 2): SGD annotated genes overexpressed in rrp6∆ strain.

Table S9 (related to Figure 2): RNA segments overexpressed in rnt1∆ strain.

Table S10 (related to Figure 2): Top 30 overexpressed regions in rnt1∆ strain.

Table S11 (related to Figure 3): RNA degradation targets predicted by “Cut and Chip”. 176

Table S12 (related to Figure 3): Top 30 cleavage targets predicted by “Cut and Chip”.

Table S13 (related to Figure 4): Sequence reads enriched in Rnt1p treated RNA.

Table S14 (related to Figure 4): Cleavage targets predicted by SALI.

Table S15 (related to Figure 4): Top 30 enriched read clusters predicted by SALI.

Table S16 (related to Figure 5): Quantitative RT-PCR measurements of Rnt1p target's expression in different growth conditions (Relative expression upon RNT1 deletion (log2)).

Table S17 (related to Figures 2, 3 and 4): GO slim terms associated with Rnt1p dependent genes.

Table S18 (related to Figures 2, 3 and 4): Gene ontology terms associated with Rnt1p dependent genes.

Table S19 (related to Figures 1 to 4): Summary of Rnt1p Dependent Targets Identified by the Different Biochemical and Genetic Techniques

177

CHAPITRE IV

Les produits de clivage d’un ARN messager rapporteur par Rnt1p sont exportés et dégradés au cytoplasme

AVANT PROPOS

Reporter mRNAs cleaved by Rnt1p are exported and degraded in the cytoplasm

Stacie Meaux, Mathieu Lavoie, Jules Gagnon, Sherif Abou Elela et Ambro van Hoof

Article publié dans

Nucleic Acids Research, volume 39, numéro 21, pages 9357-9367, 2011.

Contribution : J’ai effectué les expériences présentées à la Figure 3. Stacie Meaux a réalisé toutes les autres expériences. Jules Gagnon a contribué à la conception des constructions génétiques. J’ai participé à la rédaction des sections du manuscrit correspondantes aux résultats obtenus à la Figure 3. J’ai également contribué aux corrections subséquentes du manuscrit.

178

RÉSUMÉ

Rnt1p participe à la maturation des ARNs nucléolaires en effectuant une coupure en aval ou en amont des transcrits matures. Les produits de clivage ainsi générés présentent une extrémité 5’-monophosphate qui est alors accessible pour la dégradation par l’exoribonucléase 5’3’ Rat1p. Similairement, il a été démontré que le clivage par Rnt1p dans les régions 3’UTR de certains ARNm peut initier la terminaison de la transcription d’une manière dépendante de Rat1p. Il n’est pas clair si le clivage par Rnt1p en soi est suffisant pour amener la dégradation nucléaire des produits de clivage et la terminaison de la transcription ou si ce mécanisme est limité à quelques transcrits ayant des propriétés spécifiques. Afin de mieux comprendre le destin des ARNs clivés au noyau par Rnt1p, différentes structures tige-boucle provenant de substrats connus ont été insérées dans un gène rapporteur. Étonnamment, certains ARNm rapporteurs n’ont pas été clivés par Rnt1p. Cette absence de clivage in vivo reflète la préférence intrinsèque de Rnt1p pour certains signaux de clivage in vitro. Le clivage par Rnt1p in vivo des ARNm rapporteurs génère deux produits de clivage instables qui sont dégradés au cytoplasme par l’exosome et l’exoribonucléase 5’-3’ Xrn1p. Ce résultat suggère que : 1- le clivage par Rnt1p n’est pas suffisant pour initier la terminaison de la transcription et, 2- les produits sont rapidement exportés au cytoplasme. Ainsi, la terminaison de la transcription et la dégradation nucléaire des produits de clivage de Rnt1p ne constituent pas les mécanismes par défaut utilisés par la cellule.

179

ARTICLE 4 :

Reporter mRNAs cleaved by Rnt1p are exported and degraded in the cytoplasm

Stacie Meaux1, Mathieu Lavoie2, Jules Gagnon2, Sherif Abou Elela2,

and Ambro van Hoof1

1Departement of Microbiology and Molecular Genetics, University of Texas Health Science Center-Houston, Houston, TX 77030 and 2Département de Microbiologie et d’Infectiologie, Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec, Canada J1H 5N4

Received July 23, 2010; Revised June 28, 2011; Accepted July 15, 2011

180

Abstract

For most protein coding genes, termination of transcription by RNA polymerase II is preceded by an endonucleolytic cleavage of the nascent transcript. The 3’ product of this cleavage is rapidly degraded via the 5’ exoribonuclease Rat1p which is thought to destabilize the RNA polymerase II complex. It is not clear whether RNA cleavage is sufficient to trigger nuclear RNA degradation and transcription termination or whether the fate of the RNA depends on additional elements. For most mRNAs, this cleavage is mediated by the cleavage and polyadenylation machinery, but it can also be mediated by Rnt1p. We show that Rnt1p cleavage of an mRNA is not sufficient to trigger nuclear degradation or transcription termination. Insertion of a Rnt1p target site into a reporter mRNA did not block transcription downstream of the cleavage site, but instead produced two unstable cleavage products, neither of which were stabilized by inactivation of Rat1p. In contrast, the 3’ and 5’ cleavage products were stabilized by the deletion of the cytoplasmic 5’ exoribonuclease (Xrn1p) or by inactivation of the cytoplasmic RNA exosome. These data indicate that transcription termination and nuclear degradation is not the default fate of cleaved RNAs and that specific promoter and/or sequence elements are required to determine the fate of the cleavage products.

181

Introduction

Processing of the nascent RNA polymerase II transcript by the cleavage and polyadenylation machinery generates an uncapped 3’ cleavage product. In the torpedo model for transcription termination, degradation of this nascent uncapped transcript by Rat1p triggers transcription termination (Figure 1A) (1,2). More recently, a variant of the torpedo model of transcription termination which involves the yeast endoribonuclease III Rnt1p, has been described (3,4). Cleavage sites for the endoribonuclease Rnt1p are found downstream of the canonical polyadenylation sites of many genes and Rnt1p cleavage is thought to act as a surveillance mechanism preventing transcription read-through from leaky or inefﬁcient transcription termination (3,4). For example, cleavage of yeast NPL3 transcripts is thought to occur co-transcriptionally, generating an uncapped nascent transcript. Degradation of this nascent transcript by Rat1p triggers transcription termination thus preventing the accumulation of aberrant dicistronic mRNAs. Rnt1p cleavage has also been reported within introns of pre-mRNAs and within mRNAs (5–7). It is presumed that cleavage within these transcripts targets them for rapid decay, but it remains possible that Rnt1p cleavage mediates transcription termination. Indeed, the exact decay pathway of such products remains largely unexplored.

Interestingly, Rnt1p-mediated cleavage does not always trigger transcription termination. For example, Rnt1p is also implicated in the processing of a variety of small stable RNAs, such as snRNAs and snoRNAs (8–11). Rnt1p-mediated cleavage of the 3’ extended small RNA precursors is typically followed by further maturation by a 3’ exoribonuclease complex named the nuclear exosome (Figure 1B) (12). On the other hand, Rnt1p-mediated cleavage at speciﬁc sites in the 5’ extended snoRNA precursors generates an uncapped 5’ end that is accessible for further processing by the nuclear 5’ to 3’ exoribonuclease Rat1p (13). In some cases, multiple snoRNAs precursors are transcribed as a multimeric transcript (Figure 1C) (6,7) that is cleaved by Rnt1p resulting in monomeric intermediates which are further processed by Rat1p (14) and the nuclear exosome (15). Thus, it 182

is clear that at least for a subset of substrates, Rat1p-mediated degradation of Rnt1p cleavage products does not lead to transcription termination. However, it is still not clear if Rnt1p cleavage normally triggers transcription termination and special features or RNA binding proteins prevent Rat1p-dependent termination in the case of snoRNAs. Another possibility is that Rnt1p cleavage only initiates transcription termination if other determinants are also present.

It is likely that the fate of RNA transcripts generated by endoribonucleolytic cleavage in the nucleus is dependent on the nature of the cleavage site and adjacent sequence. Insertion of a hammerhead ribozyme or a modiﬁed group I intron, which mediates internal cleavage of a reporter mRNA, generated cleavage products that are rapidly degraded by the cytoplasmic exoribonucleases (16). These are surprising observations since ribozyme-mediated cleavage is expected to occur rapidly while the RNA is still in the nucleus which is thought to have a tight surveillance mechanism that would degrade such cleaved RNAs. Instead, it seems that ribozyme-cleaved mRNAs are exported to the cytoplasm where they are rapidly targeted by the surveillance pathways. However, it is also possible that degradation of cleaved mRNAs in the cytoplasm arise from ribozyme cleavage after nuclear export. It is also possible that nuclear RNA degradation is prevented by the structure or chemical properties of the ribozyme cleavage products that render them less susceptible to nuclear ribonucleases. For example, the 5’-OH produced by hammerhead ribozyme cleavage should generate a poor Rat1p substrate (17).

In this study, we inserted different Rnt1p cleavage signals into an mRNA. Interestingly, cleaved mRNAs were detected for some Rnt1p signals, but not for others, and accumulation of cleaved reporter mRNAs reﬂects the inherent cleavage preference of the enzyme. This provides in vivo evidence that authentic Rnt1p signals are recognized with different efﬁciencies. In the reporter mRNA context, Rnt1p-mediated cleavage does not result in transcription termination or nuclear RNA degradation. Instead, the resulting 5’ cleavage products are rapidly degraded by the cytoplasmic exosome, while the 3’ cleavage products are rapidly 183

degraded by the cytoplasmic Xrn1p. Thus, in comparable reporter mRNAs, the fate of Rnt1p cleavage products is very similar to the fate of ribozyme generated cleavage products. These results suggest that Rnt1p-mediated cleavage by itself is not sufﬁcient to trigger transcription termination or nuclear degradation. Instead, we propose that transcription termination and nuclear RNA degradation require a combination of endonucleolytic cleavage signals and special primary or secondary structural motifs.

Figure 1. The fate of different Rnt1p cleaved RNA polymerase II transcripts. (A) Transcripts from some protein-encoding genes are cleaved by Rnt1p. Cleavage is followed by processing by Rat1p, the major nuclear 5’ to 3’ exoribonuclease, which causes RNA polymerase II to dissociate from the DNA. (B) Rnt1p cleaves precursors to some snoRNAs and snRNAs at a site downstream of the mature RNA. The cleavage product is further processed by the nuclear exosome. (C) Rnt1p cleaves precursors to other snoRNAs at a site upstream or downstream of the mature RNA, or at sites both upstream and downstream. Cleavage at upstream sites is followed by processing by Rat1p, but unlike in panel A this does not lead to transcription termination. (D) To determine which of these fates is the default fate of Rnt1p cleavage products, Rnt1p cleavage stem-loops were inserted between the coding regions of Protein A and GFP. Upon cleavage of the transcript by Rnt1p, two cleavage products will result. The 5’ cleavage product will possess a 5’ cap and a 3’ hydroxyl in the place of a poly(A) tail. The 3’ cleavage product will possess a poly(A) tail and a 5’ monophosphate in the place of the 5’ cap. 184

Materials and Methods

Yeast strains

The rnt1∆ strain (yAV1084; MATa, leu2-∆0, lys2-∆0, met15∆0, ura3-∆0, his3-∆1, rnt1∆::KANMX6) was constructed by sporulating the rnt1∆::KANMX6 heterozygous diploid obtained from Open Biosystems. ski7∆, xrn1∆ and rrp6∆ strains are isogenic to the wild-type BY4741 and were obtained from Open Biosystems. The temperature-sensitive rat1-1 strain (yRP1781) was obtained from Dr Patricia J. Hilleren and Dr Roy Parker and has been described previously (18).

Plasmids

To analyze whether cleavage depended on the catalytic activity of Rnt1p, plasmids containing a wild-type allele of RNT1 [pRS315/GFP/RNT1 (19)], a catalytically- inactive allele of RNT1 [pRS315/GFP/RNT1-D247R (19)], or an empty vector were introduced into an rnt1∆ strain.

Reporter mRNA constructs were generated by ligating complementary oligonucleotides into the BamHI site of pAV214 (16). The sites where Rnt1p cleaves are indicated with / marks in the sequences below (5,9,20–22). Note that the sequences of the expected cleavage products are identical except the 3’ most 3–9 nt of the 5’ product and the 5’ most 6–8 nt of the 3’ product. For example the expected 3’ product of the MIG2 site starts with GgaaTtc, while the corresponding product of the U5 site starts with TAgaaCtc (differences in upper case). pAV351 (ProteinA-MIG2-GFP) contains the sequence 5’GGATCCAAGAAAG/CTAGCGTACAGAAACAGGAGTTTTTGACAGTAAGC/GG AATTCGGATCC3’. pAV354 (ProteinA-U5-GFP) contains the sequence 5’GGATCCTTTTCTAC/A/AATTTGATTCATGAGTCCATGGAACAAATAT/A/TAGA ACTCGGATCC3’. pAV355 (ProteinA-5U-GFP) contains the sequence 5’GGATCCGAGTTCTATATATTTGTTCCATGGACTCATGAATCAAATTTGTAGAA AAGGATCC3’. pAV356 (ProteinA-U5*-GFP) contains the sequence 5’GGATCCTTTTCTACAAATTTGATTCATGACTCGATGGAACAAATATATAGAAC 185

TCGGATCC3’. pAV398 (ProteinA-FIT2-GFP) contains the sequence 5’GGATCCATCACTTGT/ATAATTTGTAATCGAGTTCGGATAAGATGTATAC/GAA TCTGGATCC3’. pAV399 (ProteinA-snR47-GFP) contains the sequence 5’GGATCCAGA/A/AGGATATTGAACATGTTCAGAAGAACGTGGTATCTTTT/TCC TTATGGATCC3’. pAV400 (ProteinA-rRNA-GFP) contains the sequence 5’GGATCCTTCTTTC/TAAGTGGGTACTGGCAGGAGCCGGGGCCTAGTTTA/GA GAGAAGGATCC30.

In vivo RNA analysis

Yeast strains were grown in SC-URA+2% galactose to mid-log phase. The deletion strains were grown at 30°C, while the rat1-1 strain was grown at room temperature and incubated for 1 h at the restrictive temperature of 37°C. Wild-type control strains were analyzed under both conditions (e.g. in Figure 4A and C). For RNA decay analysis, the media was replaced with SC-URA+4% glucose, total RNA was extracted, separated on 1.3% agarose gel and blotted by standard methods. Blots were probed with 32P 5’ end-labeled oligonucleotides speciﬁc for Protein A (5’TCTACTTTCGGCGCCTGAGCATCATTTAGC3’), GFP (5’GCTGTTACAAACTCAAGAAGGACCATGTGG3’), RPL41A (5’GACATTACGATACTCTTGAAAGAA3’), or the RNA subunit of the signal recognition particle (SRP) as a loading control (5’GTCTAGCCGCGAGGAAGG3’). Bands were detected and quantiﬁed using a STORM phosphorImager (GE Healthcare).

Protein analysis

Western blot analysis was performed according to standard techniques, using antibodies for Protein A (Sigma) and Pgk1p (Molecular Probes).

In vitro cleavage

Recombinant Rnt1p was produced in bacteria and FLPC-puriﬁed as previously described (23). Total RNA was isolated from rnt1∆ strains containing each plasmid grown to mid-log phase in SC-URA+2% galactose. Fifteen 186

micrograms of total RNA was then incubated in the absence or presence of 10 or 80 nM recombinant Rnt1p for 10 min at 30°C in 50 ml reaction buffer [30 mM Tris– HCl (pH 7.5), 5 mM spermidine, 0.1 mM DTT, 0.1 mM EDTA (pH 7.5), 10 mM

MgCl2] supplemented or not with 150 mM KCl. The reaction was stopped by phenol: chloroform extraction and analyzed as described in the ‘In vivo RNA Analysis’ section.

For cleavage of short stem-loops, RNA transcripts were generated by T7 RNA polymerase, gel puriﬁed and 5’-end-labeled using [γ-32P]ATP as previously described (24). Cleavage reactions were performed by incubating 30 nM Rnt1p with 0.15 nM of 5’-end-labelled substrates for 10 min at 30°C in 20 ml reaction buffer, supplemented or not with 150 mM KCl, supplemented or not with 900 nM of unlabeled substrate. Cleavage products were separated by 20% denaturing PAGE and quantiﬁed as described (24).

Results

Reporter design

Different Rnt1p signals were inserted into the middle of reporter genes expressed under the control of a galactose-inducible promoter (Figure 1D). The reporter gene (16) contains the coding region of the Staphylococcus aureus protein A ZZ domain followed by the coding region of the green fluorescent protein (GFP) gene and 3’ flanking sequences of the ADH1 gene that include a canonical transcription termination/polyadenylation site. Rnt1p normally cleaves a stem-loop structure capped with an AGNN tetraloop and the cleavage efficiency is dependent on the integrity and structure of the stem-loop motif. Therefore, we tested the impact of a variety of stem-loops originating from the U5 snRNA (20), snR47 snoRNA(9),mRNA [MIG2; (5) and FIT2; (22)], and the pre-rRNA 3’ external transcribed spacer (3’ETS; 21). Except in the case of FIT2, each of these signals has been shown to trigger Rnt1p-mediated cleavage in their natural context. The 187

sequences used include the tetraloop and 22-23 nt from the natural substrates on each side of the tetraloop. The insertion of these sequences into the reporter construct preserves the coding frame and does not add any premature stop codons. Mfold (25) was used to conﬁrm that the intended stem-loop is predicted to be the most energetically stable structure (data not shown).

Rnt1p cleaves reporter mRNAs in vivo

To evaluate the cleavage efficiency of the reporter transcripts by Rnt1p, each reporter plasmid was transformed into a wild-type yeast strain and into the isogenic strain carrying a deletion of the RNT1 gene. After galactose-induced expression of the reporter mRNA, total RNA was extracted from the transformed strains and analyzed by northern blotting. ProteinA-GFP reporters lacking an Rnt1p cleavage signal or containing a hammerhead ribozyme were included to serve as controls. As shown in Figure 2A, hybridization with a probe specific for the 5’ half of the reporter (Protein A) revealed two distinct mRNA species. The slower migrating species corresponds to the expected size of the uncleaved reporter transcript, indicating that all reporters are effectively transcribed. More importantly, a faster migrating fragment corresponding to the expected size of the cleaved 5’ portion of the reporter is observed in the wild-type strain transformed with the constructs containing the ribozyme control, U5 or rRNA cleavage signal. Cleaved reporter mRNAs with U5- or rRNA-derived sequences were not detected in an isogenic rnt1∆ strain, suggesting that the smaller species were indeed cleavage products produced by Rnt1p. Similarly, 3’ cleavage products (GFP) were detectable for these same constructs in the wild-type strain, but not in an rnt1∆ strain (Figure 2A). Interestingly, both 5’ and 3’ cleavage products can be observed at the same time for the cleaved construct, thus suggesting that Rnt1p cleavage was not sufficient to prevent further transcription of the downstream GFP fragment. As expected, deleting RNT1 had no effect on the RNAs produced from the ribozyme control (data not shown). We also detected low levels of smaller RNAs from the construct 188

with the FIT2-derived Rnt1p site, but these RNAs could also be detected in an rnt1∆ strain, suggesting that they do not result from Rnt1p-mediated cleavage. For the constructs with Rnt1p sites derived from snR47 and MIG2, we failed to detect any putative cleavage products. Thus it appears that some, but perhaps not all, of our reporter mRNAs can be efﬁciently cleaved by Rnt1p in vivo.

To further show that the ProteinA-U5-GFP construct was an Rnt1p substrate, we altered the stem-loop sequence. In one control, the AGNN loop sequence and the top base pair of the stem were changed (GAGTCC changed to GACTCG; Figure 2B; construct ProteinA–U5*–GFP), and in a second control we inserted the U5-derived sequence backwards (Figure 2B; construct ProteinA- 5U- GFP). Neither of these controls yielded any cleavage product, consistent with the known speciﬁcity of Rnt1p. Finally, the D245R point mutation in the catalytic site of Rnt1p (19) also disrupted the accumulation of the cleavage product (Figure 2C). Overall, we conclude that the reporter mRNAs with U5 snRNA and rRNA derived sequences are cleaved by Rnt1p in vivo and that this cleavage is dependent upon the AGNN stem-loop.

189

Figure 2. Rnt1p cleaves reporter mRNAs in vivo. (A) Northern blot analysis of total RNA extracted from wild-type or isogenic rnt1∆ yeast strains transformed with the indicated plasmids. (B) Northern blot analysis of total RNA extracted from wild-type yeast transformed with the Protein A- U5-GFP plasmid and two variants that should not be recognized by Rnt1p: Protein A-U5*-GFP contains point mutations of the U5 sequence that disrupt Rnt1p recognition and Protein A-5U-GFP contains the U5-derived sequence in the inverted orientation. (C) Northern blot analysis of RNA isolated from an rnt1∆ strain transformed with the ProteinA-U5-GFP reporter construct and either an empty vector, a plasmid expressing wild- type Rnt1p, or a plasmid expressing a catalytically inactive Rnt1p (D245R).

Accumulation of cleaved reporter mRNAs in vivo reﬂects substrate preference of Rnt1p

The fact that cleavage could not be observed for three of our reporter transcripts was somewhat surprising given that these cleavage signals are all derived from known natural Rnt1p substrates. Since the potential products generated after Rnt1p cleavage would be very similar from one construct to another, we believe that failure to detect cleavage products of mRNAs with MIG2, FIT2 and snR47 stem-loops is not caused by the fast decay of these speciﬁc transcripts in vivo, although we could not completely rule out this possibility. Another possible explanation for the lack of accumulation of cleaved reporter 190

mRNAs is that Rnt1p may have an intrinsic preference for the U5 and rRNA stem- loops. To test this possibility we isolated total RNA from the rnt1∆ strain containing each of the five constructs. This total RNA was then incubated with recombinant Rnt1p purified from Escherichia coli and submitted to northern blot analysis (probing for GFP). This analysis showed that all five AGNN stem-loop mRNAs are specifically cleaved under non-physiological conditions that favor RNA-protein interactions (high concentrations of Rnt1p and low salt buffer; Figure 3A). This result indicates that all transcripts can form the expected secondary structure in vitro. Importantly, we could not detect specific cleavage products of the control lacking an AGNN stem-loop, although the uncleaved mRNA seems unspecifically degraded in the presence of Rnt1p under these conditions. Reducing the enzyme concentration and increasing KCl to physiological concentration (150 mM) suppresses the unspecific degradation of the control lacking the stem-loop and reveals that the mRNAs with Rnt1p sites from U5 snRNA and rRNA were preferentially cleaved (Figure 3B). These results signify that although all reporter transcripts have the potential to be cleaved by Rnt1p, only the U5 and rRNA reporters are cleaved under physiological conditions. Thus, the differences in cleaved reporter mRNA accumulation in vivo correlate with differences in interactions of Rnt1p with these AGNN stem-loops.

It was not clear at this point whether Rnt1p’s preferential cleavage of some substrates is a secondary effect of the insertion of the different cleavage signals in an mRNA, out of their normal context, or if Rnt1p has inherent preferences for certain AGNN stem-loops. To test if the stem-loops themselves show a difference in their ability to be cleaved by Rnt1p, we compared the cleavage of small, in vitro- transcribed U5-derived or MIG2-derived recognition sequences. We chose these two substrates because our in vivo and in vitro studies suggested that the U5 stem- loop is a good substrate for cleavage while the MIG2 stem-loop is cleaved less efﬁciently. Following in vitro transcription and 5’ end labeling, each substrate was incubated with or without puriﬁed Rnt1p, either in the absence or presence of 150 mM KCl. Interestingly, the cleavage of these short stem-loop substrates mirrored 191

those seen with the corresponding reporter mRNAs. Indeed, both substrates were cleaved in low salt concentrations when Rnt1p is in excess (enzyme:RNA ratio of 200:1). Increasing the salt concentration did not affect the cleavage efficiency of the U5 stem-loop, but strongly inhibited cleavage of the MIG2-derived substrate. Consistently, the cleavage of the MIG2 substrate was less efficient than that observed with U5 under more physiological conditions (150 mM KCl and enzyme:RNA ratio of 1:30). We conclude that different sites for Rnt1p exhibit different cleavage efficiencies and that the newly constructed in vivo reporters faithfully reflect the substrates inherent cleavage efficiency and thus constitute a good tool for testing Rnt1p cleavage in vivo.

Figure 3. Differences in the accumulation of cleaved reporter mRNAs in vivo correlate with differences in recognition by Rnt1p. (A and B) Total RNA was isolated from an rnt1∆ strain containing each of the indicated plasmids and incubated for 10 min with either high concentration (80 nM) of Rnt1p and no KCl added (A) or low concentration (10 nM) of Rnt1p and 150 mM KCl (B). Each sample was then submitted to Northern blot analysis and probed for GFP. (C) Short RNAs containing Rnt1p cleavage sites from U5 or MIG2 were generated by T7 RNA polymerase and 5’-end- labeled. Substrates were incubated with or without puriﬁed Rnt1p, either in the presence or absence of potassium chloride as indicated.

192

The 5’ product of an Rnt1p cleaved reporter mRNA is degraded by cytoplasmic RNases

It was previously shown that mRNAs cleaved by a hammerhead ribozyme are degraded by cytoplasmic enzymes (16,26), suggesting that the unadenylated 5’ cleavage product is exported from the nucleus. However, we could not exclude the possibility that the cleavage by the ribozyme occurred in the cytoplasm after export of the full-length capped and polyadenylated mRNA from the nucleus. Given the nuclear localization of Rnt1p (19,27,28), our newly designed ProteinA–U5–GFP construct now allows us to evaluate the fate of mRNAs that are cleaved in the nucleus.

In order to identify the cellular location where the cleavage products are degraded, we tested the impact of impairing nuclear or cytoplasmic exoribonucleases on the stability of Rnt1p cleavage products. If an Rnt1p cleavage product is degraded in the nucleus, it should be stabilized by the inactivation of the nuclear 5’ exoribonuclease Rat1p and/or the nuclear exosome. On the other hand, if the cleaved RNA is degraded in the cytoplasm it is expected to be more stable upon inactivation of the cytoplasmic 5’ exoribonuclease Xrn1p and/or the cytoplasmic exosome. As shown in Figure 4, the Rnt1p 5’-end cleavage product was equally unstable in wild-type cells, cells carrying a temperature-sensitive mutation in RAT1 and grown at the restrictive temperature, or cells lacking the nuclear exosome component Rrp6p. In contrast, in strains lacking the cytoplasmic exosome cofactor Ski7p or the cytoplasmic exonuclease Xrn1p, the half-life of the 5’ cleavage product is increased 4- and 2-fold, respectively (Figure 4), suggesting that the 5’ cleavage product is exported to the cytoplasm. If Rnt1p 5’ cleavage product is exported to the cytoplasm, it should be available to the translational machinery. Western blot analysis using antibodies against Protein A indicated that the unadenylated cleavage product is translated in wild-type cells, and the amount of the protein produced is increased in ski7∆ cells, where the 5’ cleavage product is more stable (Figure 4D). Thus, both mRNA stability measurements and western 193

blot analysis indicates that the Rnt1p 5’ cleavage product is exported to the cytoplasm.

We have previously shown that introducing a ribozyme motif within the reporter constructs generates an unadenylated cleavage product that is exported and translated (16,29). To determine the polyadenylation state of Rnt1p cleavage products we used oligo(dT)-dependent RNase H digestion. As expected, the vast majority of the 5’ product of Rnt1p cleavage was not polyadenylated (Figure 4E). We conclude that the Rnt1p 5’ cleavage product is predominantly degraded by cytoplasmic exoribonucleases and suggest that Rnt1p-cleaved mRNAs are not retained in the nucleus for an extended period of time, but instead are readily exported from the nucleus.

In addition to the cleavage products, we also detected a signiﬁcant amount of uncleaved ProteinA-U5-GFP mRNA. This uncleaved mRNA pool is stabilized in an xrn1∆ strain, but not in strains with mutations in RRP6, RAT1 or SKI7. This Xrn1p-mediated decay suggests that the uncleaved mRNA represents a pool of mRNA that escapes Rnt1p cleavage and is exported from the nucleus as a typical mRNA. Consistent with this interpretation is that mutation of the Rnt1p-recognition site does not signiﬁcantly stabilize the uncleaved mRNA (data not shown) and that the uncleaved mRNA is translated into a protein A-GFP fusion protein (Figure 4D). Together these observations indicate that the uncleaved pool of ProteinA-U5-GFP mRNA that is detected by northern blot is not a precursor to the cleaved pool, but instead is degraded like a normal mRNA that lacks an Rnt1p cleavage site.

194

Figure 4. The unadenylated 5’ cleavage product is degraded by cytoplasmic RNases. (A) Isogenic wild-type, ski7∆, rrp6∆ or xrn1∆ strains containing the Protein A-U5-GFP reporter were grown to mid-log phase at 30°C. Transcription of the reporter gene was terminated by addition of glucose and total RNA was isolated at the indicated times. RNA was submitted to northern blot analysis and probed for Protein A or SRP RNA. Half-lives for uncleaved and cleaved RNAs with standard deviations are indicated. Each experiment was performed four times. (B) Bar graph of the 195

half-lives of the Protein A cleavage product in mutants tested in panel A, with standard deviations. Asterisk indicates that the average half-life is signiﬁcantly different from the half-life in wild-type (P < 0.05). (C) Wild-type yeast or a yeast strain with a temperature-sensitive allele of RAT1, both expressing the Protein A-U5-GFP reporter, were grown to mid-log phase in media containing galactose at 23°C and then incubated at 37°C for 1 h to inactivate the Rat1-1p. Media was then replaced with media containing glucose to terminate transcription of the reporter and aliquots were taken at the times indicated. Decay rates for uncleaved and cleaved RNAs are indicated. (D) The 5’ cleavage product is translated. Isogenic wild-type and ski7∆ strains were transformed with empty vector or the Protein A-U5-GFP plasmid as indicated. Total protein and RNA was isolated from a culture grown to mid-log phase and submitted to western and northern blot analysis, respectively. Western blots were probed for Protein A and Pgk1p, to control for loading, while northern blots were probed for Protein A and the RNA subunit of the signal recognition particle (SRP). (E) The 5’ cleavage product is unadenylated. Shown is a polyacrylamide northern blot of total RNA extracted from wild-type yeast transformed with the indicated plasmids. The RNA was treated with RNase H in the presence or absence of oligo(dT) as indicated. The shift in mobility for the RPL41A mRNA indicates that it contains a poly(A) tail that is removed by RNase H treatment, while the lack of a shift in mobility of the Protein A mRNA and the RNA subunit of the signal recognition particle (SRP) indicates that these RNAs do not have a poly(A) tail. For each panel, the indicated strains were grown to mid-log phase at 23°C prior to total RNA isolation. The blots were hybridized with probes speciﬁc for Protein A, RPL41A and the RNA subunit of the SRP as a loading control as indicated.

The 3’ product of an Rnt1p cleaved reporter mRNA is degraded by Xrn1p

When analyzing the fate of the 3’ GFP cleavage product, we discovered that, in a wild-type background, this product was present at extremely low levels (Figure 5A) and extremely unstable (Figure 5B). In the xrn1∆ strain, steady state levels were increased ~18-fold compared to wild-type (Figure 5A), and its half-life was greatly increased (half-life of 22 min; Figure 5D). In contrast, the stability was not affected by inactivation of the nuclear 5’–3’ exoribonuclease Rat1p (Figure 5C), or by inactivation of the nuclear or cytoplasmic exosome (data not shown). These results suggest that Xrn1p is mainly responsible for the decay of the 3’ cleavage 196

product, and therefore that it is not retained in the nucleus for an extended period of time, but instead readily exported from the nucleus.

Figure 5. The uncapped 3’ cleavage product is degraded by the cytoplasmic Xrn1p. (A) Isogenic wild-type and xrn1∆ strains containing the Protein A-U5-GFP reporter were grown to mid-log phase in media containing galactose and submitted to northern blot analysis. Blots were probed for GFP and the RNA subunit of the signal recognition particle (SRP) as indicated. (B–D) The degradation rate of cleaved GFP mRNA was determined in wild-type, xrn1∆ and rat1-1 strains as described in Figure 4.

197

Discussion

Many examples in the literature appear to indicate that Rnt1p-mediated cleavage during non-coding RNA processing is more efficient than cleavage of stem-loops in mRNAs. However, it is difficult to directly compare Rnt1p cleavage efficiency between substrates in their natural context (between snRNA U5 and MIG2 mRNA for example) because many variables can affect substrate recognition, enzyme catalysis, and fate of the cleavage products. In addition, Rnt1p cleavage of some substrates may be influenced by features outside the stem-loop itself, such as interacting proteins or specific RNA folding (11,30). Introducing different Rnt1p cleavage sites derived from varying natural substrates into a common mRNA context provides a useful tool to study Rnt1p-mediated cleavage and the fate of mRNA cleavage products that are generated in the nucleus.

Interestingly, not all stem-loops resulted in the accumulation of cleaved reporter mRNAs. We have shown that this is at least in part because the inserted U5 stem-loop is recognized much more efficiently than the MIG2 signal. In vitro experiments confirmed that this preferential cleavage is caused by specific features within the AGNN stem-loops themselves. One of many possible explanations is the presence of a GC-rich region near the cleavage site of the MIG2 stem. Such regions can negatively affect Rnt1p cleavage of short substrates in vitro (24), but it was not known if this effect would also apply in vivo. More importantly, an independent study published during the writing of this manuscript has reported similar conclusions (31). Basically, it was shown that Rnt1p activity in vivo was highly variable when synthetic hairpins were inserted into the 3’ untranslated portion of a transcript and that in vivo activity could be correlated with in vitro cleavage rate. Babiskin and Smolke (31) showed that most of their synthetic hairpins selected for efficient cleavage in one reporter context maintained their cleavage efficiency in a second reporter context. Our results suggest that MIG2, FIT2 and snR47 stem-loops are not well cleaved outside of their natural context. At least for snR47, cleavage of the endogenous substrate is efficient, with little if any pre-snR47 detectable in cells containing a functional Rnt1p (9). Thus, the MIG2, 198

FIT2 and snR47 stem-loops may not fold properly in the context of a reporter, or cleavage of these natural stem-loops (as opposed to selected synthetic hairpins) requires additional features not present in reporter constructs. Finally, although unlikely, we have not been able to rule out that these reporter mRNAs are in fact cleaved, but that the cleavage products are very rapidly degraded by one or more RNases.

In addition to constituting a convenient tool to evaluate the elements that control Rnt1p activity, our reporter system allows us to study the impact of cleavage on gene expression. For example, it was shown that Rnt1p could mediate transcription termination of transcripts (such as NPL3) in the absence of a poly-A signal downstream of the ORF, thus preventing the accumulation of a polycistronic transcript. Similarly, one might expect that cleavage of the reporter transcript would terminate transcription and prevent expression of the downstream sequence (GFP). Clearly our results indicate that this is not the case. Although we cannot tell if cleavage occurred co- or post-transcriptionally, the fact that both 5’ and 3’ cleavage products can be observed at the same time (Figure 2A) indicates that an Rnt1p target site is not sufﬁcient for transcription termination. Moreover, if Rnt1p cleavage triggered Rat1p-dependent transcription termination, one would expect to observe in a rat1-1 mutant strain higher expression levels of the 3’ cleavage product, without an effect on the stability of the 3’ cleavage product. In contrast to this expectation, there was no large increase in the levels of the 3’ cleavage product in a rat1-1 strain (Figure 5). We conclude that cleavage of our reporter constructs did not lead to signiﬁcant levels of transcription termination.

This conclusion raises the question of what features of Rnt1p cleavage trigger or prevent transcription termination. Our data indicate that different Rnt1p sites are cleaved with different efﬁciencies. One possible hypothesis could be that fast Rnt1p cleavage is required for co-transcriptional cleavage and thus for triggering transcription termination. Interestingly, the U5 snRNA and rRNA sites that were efﬁciently cleaved in our reporter mRNAs, are naturally present 3’ of the mature RNA and cleavage at the endogenous rRNA site facilitates transcription 199

termination (32,33). In contrast, slow Rnt1p cleavage would rather result in post- transcriptional cleavage. Consistently, the natural position for Rnt1p site in snR47 is 5’ of the mature snoRNA, where it facilitates processing, but not transcription termination. If this site triggered transcription termination by the torpedo model it would prevent snoRNA production. Importantly, our results suggest that the required differences in timing of cleavage are in part determined by Rnt1p site itself (Figure 3C). Some Rnt1 target sites may be cleaved rapidly, allowing for the cleavage to trigger transcription termination at the 3’ end of genes. Conversely, Rnt1p target sites upstream of mature snoRNAs may need to be cleaved slower, allowing for continued transcription and processing. In vivo Rnt1p-mediated cleavage could be further reduced by alternative RNA or RNP structures, or increased by recruiting Rnt1p to the site. RNA polymerase II pausing after transcribing an Rnt1p site could also result in increased co-transcriptional RNA cleavage, and increased transcription termination. Thus, endogenous genes that undergo Rnt1p-mediated transcription termination may contain additional sequences outside the stem-loop that either cause very efﬁcient cleavage by Rnt1p, or transcriptional pausing downstream of the Rnt1p site.

The downstream events that follow Rnt1p cleavage on non-coding RNA substrates, such as rRNA, snoRNAs and snRNAs are well characterized (12,15). However, the fate of cleaved mRNAs was still obscure. In this study, we have observed that the 3’ product of Rnt1p cleavage was very unstable in wild-type cells, but signiﬁcantly stabilized upon deletion of the XRN1 gene, which encodes the major cytoplasmic 5’ exoribonuclease. In contrast, inactivation of Rat1p had no detectable effect on the stability of this fragment. Since Rnt1p has been localized to the nucleus by several independent methods (19,27,28), we conclude that our reporter mRNAs are cleaved in the nucleus, and that the 3’ fragment is degraded following export to the cytoplasm. This fate is the same as what we have previously shown for hammerhead ribozyme cleavage products, and thus likely presents a general degradation pathway for cleaved mRNAs. This implies that endogenous Rnt1p cleavage products contain speciﬁc features to specify their further 200

processing. Interestingly, the endogenous RPS22B and RPL18A mRNAs, which both encode ribosomal proteins, both contain an Rnt1p site within an intron. The cleaved RPS22B pre-mRNA accumulates in a rat1-1 strain, while the cleaved RPL18A pre-mRNA accumulates in an xrn1∆ strain. Although both accumulate to even higher levels in the double mutant, this shows that these apparently very similar Rnt1p products have different downstream fates (7).

The 5’ fragment resulting from Rnt1p cleavage of our reporter mRNAs was stabilized by mutations inactivating the cytoplasmic exosome, and to a lesser extent by deleting XRN1. This 5’ fragment also is translated. We conclude that this 5’ fragment is exported to the cytoplasm, where it is degraded. This fate resembles what we have previously described for the 5’ cleavage product of a hammerhead ribozyme, and thus likely is the default pathway for cleaved mRNAs. Since endogenous snRNAs and snoRNAs are further processed by the nuclear exosome, this suggests that they too contain speciﬁc features to direct such further processing and that Rnt1p cleavage by itself is not sufﬁcient to dictate the fate of the cleavage products.

Both the poly(A) tail and the cap structure have been suggested to facilitate mRNA export from the nucleus. Our findings confirm that neither is absolutely required for this process. Previously, we and others have shown that ribozyme- cleaved mRNAs also enter the cytoplasm (16,34,35), but an important limitation to these studies is that it is hard to prove that ribozyme cleavage occurred before nuclear export. The strategy of using Rnt1p should be widely applicable to generate unadenylated mRNAs specifically in the nucleus. Furthermore, Rnt1p cleavage products contain a 3’ OH and a 5’ monophosphate (36-38), while hammerhead ribozymes generate a 2’3’ cyclic phosphate and a 5’ OH (39,40). Thus, Rnt1p products are chemically the same as the cleavage products of most nuclear and cytoplasmic endo- and exoribonucleases. The demonstration of the utility of Rnt1p to cleave mRNAs in the nucleus complements other tools, such as the use of an rlg1-100 strain to specifically generate cleaved HAC1 mRNA in the cytoplasm (41), and the use of allosteric ribozymes to temporally regulate reporter 201

mRNA cleavage (42). Finally, our reporter construct should provide a very convenient tool in order to identify the regulatory features, such as promoter sequences or regions ﬂanking endogenous cleavage signals, which trigger Rnt1p- mediated transcription termination and dictate the fate of the cleavage products.

Acknowledgements

The authors thank Dr Patricia J. Hilleren and Dr Roy Parker for providing the rat1-1 strain.

Funding

Pew Scholarship program in the Biomedical Sciences; National Institutes of Health (GM069900 to A.v.H.); National Science Foundation (1020739 to A.v.H.); Canadian Institute of Health Research (CIHR; to S.A.). Funding for open access charge: National Science Foundation.

Conﬂict of interest statement. None declared.

202

References

1. Kim,M., Krogan,N.J., Vasiljeva,L., Rando,O.J., Nedea,E., Greenblatt,J.F. and Buratowski,S. (2004) The yeast Rat1 exonuclease promotes transcription termination by RNA polymerase II. Nature, 432, 517–522.

2. West,S., Gromak,N. and Proudfoot,N.J. (2004) Human 5’3’ exonuclease Xrn2 promotes transcription termination at co-transcriptional cleavage sites. Nature, 432, 522–525.

3. Ghazal,G., Gagnon,J., Jacques,P.E., Landry,J.R., Robert,F. and Elela,S.A. (2009) Yeast RNase III triggers polyadenylation-independent transcription termination. Mol. Cell., 36, 99–109.

4. Rondon,A.G., Mischo,H.E., Kawauchi,J. and Proudfoot,N.J. (2009) Fail-safe transcriptional termination for protein-coding genes in S. cerevisiae. Mol. Cell., 36, 88–98.

5. Ge,D., Lamontagne,B. and Elela,S.A. (2005) RNase III-mediated silencing of a glucose-dependent repressor in yeast. Curr. Biol., 15, 140–145.

6. Larose,S., Laterreur,N., Ghazal,G., Gagnon,J., Wellinger,R.J. and Elela,S.A. (2007) RNase III-dependent regulation of yeast telomerase. J. Biol. Chem., 282, 4373–4381.

7. Danin-Kreiselman,M., Lee,C.Y. and Chanfreau,G. (2003) RNAse III- mediated degradation of unspliced pre-mRNAs and lariat introns. Mol. Cell., 11, 1279–1289.

8. Abou Elela,S. and Ares,M. Jr. (1998) Depletion of yeast RNase III blocks correct U2 3’ end formation and results in polyadenylated but functional U2 snRNA. EMBO J., 17, 3738–3746.

9. Chanfreau,G., Legrain,P. and Jacquier,A. (1998) Yeast RNase III as a key processing enzyme in small nucleolar RNAs metabolism. J. Mol. Biol., 284, 975–988.

10. Seipelt,R.L., Zheng,B., Asuru,A. and Rymond,B.C. (1999) U1 snRNA is cleaved by RNase III and processed through an Sm site-dependent pathway. Nucleic Acids Res., 27, 587–595.

11. Ghazal,G., Ge,D., Gervais-Bird,J., Gagnon,J. and Abou Elela,S. (2005) Genome-wide prediction and analysis of yeast RNase III-dependent snoRNA processing signals. Mol. Cell. Biol., 25, 2981–2994.

203

12. Allmang,C., Kufel,J., Chanfreau,G., Mitchell,P., Petfalski,E. and Tollervey,D. (1999) Functions of the exosome in rRNA, snoRNA and snRNA synthesis. EMBO J., 18, 5399–5410.

13. Lee,C.Y., Lee,A. and Chanfreau,G. (2003) The roles of endonucleolytic cleavage and exonucleolytic digestion in the 5’-end processing of S. cerevisiae box C/D snoRNAs. RNA, 9, 1362–1370.

14. Petfalski,E., Dandekar,T., Henry,Y. and Tollervey,D. (1998) Processing of the precursors to small nucleolar RNAs and rRNAs requires common components. Mol. Cell. Biol., 18, 1181–1189.

15. van Hoof,A., Lennertz,P. and Parker,R. (2000) Yeast exosome mutants accumulate 3’-extended polyadenylated forms of U4 small nuclear RNA and small nucleolar RNAs. Mol. Cell. Biol., 20, 441–452.

16. Meaux,S. and van Hoof,A. (2006) Yeast transcripts cleaved by an internal ribozyme provide new insight into the role of the cap and poly(A) tail in translation and mRNA decay. RNA, 12, 1323–1337.

17. Stevens,A. and Maupin,M.K. (1987) A 5’ to 3’ exoribonuclease of Saccharomyces cerevisiae: size and novel substrate speciﬁcity. Arch. Biochem. Biophys., 252, 339–347.

18. Hilleren,P.J. and Parker,R. (2003) Cytoplasmic degradation of splice- defective pre-mRNAs and intermediates. Mol. Cell., 12, 1453–1465.

19. Catala,M., Lamontagne,B., Larose,S., Ghazal,G. and Elela,S.A. (2004) Cell cycle-dependent nuclear localization of yeast RNase III is required for efﬁcient cell division. Mol. Biol. Cell., 15, 3015–3030.

20. Chanfreau,G., Elela,S.A., Ares,M. Jr. and Guthrie,C. (1997) Alternative 3’- end processing of U5 snRNA by RNase III. Genes Dev., 11, 2741–2751.

21. Kufel,J., Dichtl,B. and Tollervey,D. (1999) Yeast Rnt1p is required for cleavage of the pre-ribosomal RNA in the 3’ ETS but not the 5’ ETS. RNA, 5, 909–917.

22. Lee,A., Henras,A.K. and Chanfreau,G. (2005) Multiple RNA surveillance pathways limit aberrant expression of iron uptake mRNAs and prevent iron toxicity in S. cerevisiae. Mol. Cell., 19, 39–51.

23. Lamontagne,B. and Elela,S.A. (2001) Puriﬁcation and characterization of Saccharomyces cerevisiae Rnt1p nuclease. Methods Enzymol., 342, 159– 167.

204

24. Lamontagne,B., Ghazal,G., Lebars,I., Yoshizawa,S., Fourmy,D. and Elela,S.A. (2003) Sequence dependence of substrate recognition and cleavage by yeast RNase III. J. Mol. Biol., 327, 985–1000.

25. Zuker,M. (2003) Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res., 31, 3406–3415.

26. Meaux,S., van Hoof,A. and Baker,K.E. (2008) Nonsense-mediated mRNA decay in yeast does not require PAB1 or a poly(A) tail. Mol. Cell., 29, 134– 140.

27. Henras,A.K., Bertrand,E. and Chanfreau,G. (2004) A cotranscriptional model for 3’-end processing of the Saccharomyces cerevisiae pre-ribosomal RNA precursor. RNA, 10, 1572–1585.

28. Huh,W.K., Falvo,J.V., Gerke,L.C., Carroll,A.S., Howson,R.W., Weissman,J.S. and O’Shea,E.K. (2003) Global analysis of protein localization in budding yeast. Nature, 425, 686–691.

29. Meaux,S. (2008) Ph.D. Thesis. UTHSC-Houston, Houston, TX.

30. Giorgi,C., Fatica,A., Nagel,R. and Bozzoni,I. (2001) Release of U18 snoRNA from its host intron requires interaction of Nop1p with the Rnt1p endonuclease. EMBO J., 20, 6856–6865.

31. Babiskin,A.H. and Smolke,C.D. (2011) A synthetic library of RNA control modules for predictable tuning of gene expression in yeast. Mol. Syst. Biol., 7, 471.

32. El Hage,A., Koper,M., Kufel,J. and Tollervey,D. (2008) Efﬁcient termination of transcription by RNA polymerase I requires the 5’ exonuclease Rat1 in yeast. Genes Dev., 22, 1069–1081.

33. Prescott,E.M., Osheim,Y.N., Jones,H.S., Alen,C.M., Roan,J.G., Reeder,R.H., Beyer,A.L. and Proudfoot,N.J. (2004) Transcriptional termination by RNA polymerase I requires the small subunit Rpa12p. Proc. Natl Acad. Sci. USA, 101, 6068–6073.

34. Dower,K., Kuperwasser,N., Merrikh,H. and Rosbash,M. (2004) A synthetic A tail rescues yeast nuclear accumulation of a ribozyme-terminated transcript. RNA, 10, 1888–1899.

35. Duvel,K., Valerius,O., Mangus,D.A., Jacobson,A. and Braus,G.H. (2002) Replacement of the yeast TRP4 3’ untranslated region by a hammerhead ribozyme results in a stable and efﬁciently exported mRNA that lacks a poly(A) tail. RNA, 8, 336–344. 205

36. Ji,X. (2006) Structural basis for non-catalytic and catalytic activities of ribonuclease III. Acta Crystallogr. D Biol. Crystallogr., 62, 933–940.

37. Li,W.M., Barnes,T. and Lee,C.H. (2010) Endoribonucleases– enzymes gaining spotlight in mRNA metabolism. FEBS J., 277, 627–641.

38. MacRae,I.J. and Doudna,J.A. (2007) Ribonuclease revisited: structural insights into ribonuclease III family enzymes. Curr. Opin. Struct. Biol., 17, 138–145.

39. Hutchins,C.J., Rathjen,P.D., Forster,A.C. and Symons,R.H. (1986) Self- cleavage of plus and minus RNA transcripts of avocado sunblotch viroid. Nucleic Acids Res., 14, 3627–3640.

40. Prody,G.A., Bakos,J.T., Buzayan,J.M., Schneider,I.R. and Bruening,G. (1986) Autolytic processing of dimeric plant virus satellite RNA. Science, 231, 1577–1580.

41. Sidrauski,C., Cox,J.S. and Walter,P. (1996) tRNA ligase is required for regulated mRNA splicing in the unfolded protein response. Cell, 87, 405– 413.

42. Win,M.N. and Smolke,C.D. (2007) A modular and extensible RNA-based gene-regulatory platform for engineering cellular function. Proc. Natl Acad. Sci. USA, 104, 14283–14288.

206

CHAPITRE V

Le couplage de la répression de la transcription et la dégradation de l’ARN permet la régulation de l’expression conditionnelle des gènes

AVANT PROPOS

Regulation of conditional gene expression by coupled transcription repression and RNA degradation

Mathieu Lavoie, Dongling Ge et Sherif Abou Elela

Article publié dans

Nucleic Acids Research, volume 40, numéro 2, pages 871-83, 2012

Contribution : J’ai réalisé toutes les expériences présentées à l’exception des essais d’extension d’amorces de la Figure 1 qui ont été réalisées par Dongling Ge. Dongling Ge a aussi préparé les souches de levures ainsi que les plasmides. J’ai préparé toutes les figures et rédigé les légendes des figures et les « matériels et méthodes ». J’ai écrit une version initiale des sections « résultats » et « discussion » auxquelles Sherif Abou Elela a apporté plusieurs modifications. J’ai également contribué à toutes les révisions subséquentes du manuscrit.

207

RÉSUMÉ

La présence de glucose dans l’environnement de la levure Saccharomyces cerevisiae active des voies de signalisation qui provoquent d’importants changements de l’activité transcriptionnelle. D’un autre côté, de nombreuses études suggèrent que la régulation post-transcriptionnelle est également importante pour la réponse au glucose. Par exemple, il a été démontré que Rnt1p régule l’expression de l’ARNm codant pour Mig2p, un répresseur de transcription impliqué dans la voie de signalisation du glucose. Dans cette étude, nous démontrons que la dégradation de l’ARN régule l’expression de gènes associés avec la voie de signalisation du glucose Snf3p-Rgt2p en combinaison avec la répression transcriptionnelle. En effet, les résultats montrent que la délétion du gène RNT1 augmente l’expression des facteurs de transcription Mth1 et Rgt1 in vivo et que Rnt1p clive leurs ARNm in vitro. Notamment, Rnt1p participe à la répression dépendante du glucose de Mth1, contrairement à Rgt1 dont l’expression n’est pas influencée par les conditions de culture. Ces résultats suggèrent que Rnt1p peut réguler l’expression des gènes de manière constitutive ou conditionnelle. L’impact de Rnt1p sur la répression de Mig2 et de Mth1 est partiellement dépendant de la région promotrice des gènes. De plus, la délétion de RNT1 active la transcription d’un gène rapporteur LacZ placé en aval des régions promotrices des gènes MIG2, RGT1 et MTH1, indiquant que Rnt1p régule la transcription de ces gènes indépendamment de la présence d’un signal de clivage. Au final, les résultats suggèrent un modèle dans lequel Rnt1p est recruté au niveau des régions promotrices des gènes de la voie Snf3p-Rgt2p et clive co- transcriptionnellement leur ARNm afin d’assurer une répression rapide de l’expression génique en réponse aux fluctuations de nutriments dans le milieu.

208

ARTICLE 5 :

Regulation of conditional gene expression by coupled transcription repression and RNA degradation

Mathieu Lavoie, Dongling Ge and Sherif Abou Elela

RNA Group, Département de microbiologie et d’infectiologie, Faculté de médecine

et des sciences de la santé, Université de Sherbrooke, Sherbrooke, Québec,

Canada, J1H 5N4.

Received June 1, 2011; Revised and Accepted August 30, 2011

209

Abstract

Gene expression is determined by a combination of transcriptional and posttranscriptional regulatory events that were thought to occur independently. This report demonstrates that the genes associated with the Snf3p-Rgt2p glucose- sensing pathway are regulated by interconnected transcription repression and RNA degradation. Deletion of the dsRNA specific ribonuclease III Rnt1p increased the expression of Snf3p-Rgt2p associated transcription factors in vivo and degraded their messenger RNA (mRNA) in vitro. Surprisingly, Rnt1p’s effect on gene expression in vivo was both RNA and promoter dependent, thus linking RNA degradation to transcription. Strikingly, deletion of RNT1 induced promoter specific transcription of the glucose sensing genes even in the absence of RNA cleavage signals. Together, the results presented here support a model in which cotranscriptional RNA degradation increases the efficiency of gene repression, thereby allowing an effective cellular response to the continuous changes in nutrient concentrations.

210

Introduction

In higher eukaryotes, conditional mRNA degradation is believed to be generally initiated by the dsRNA specific ribonuclease Dicer (1). Dicer cleavage generates either short interfering RNA (siRNA), or microRNA (miRNA), which triggers an RNA interference (RNAi) pathway that leads to complete degradation of the targeted mRNA (2-6). Sequence complementarity between the Dicer products and the targeted mRNA determines the site of cleavage, and confers high specificity to this RNA degradation strategy. RNAi dependent mRNA degradation has been identified in most eukaryotes including the fission yeast Schizosaccharomyces pombe (7-9). Saccharomyces cerevisiae is among the few eukaryotes that do not express the known components of the RNAi machinery. Instead, budding yeast express only a single isoform of RNase III (Rnt1p) that is required for the maturation of both pre-rRNA and snoRNA (10,11). Recently, this enzyme was also shown to initiate the degradation of several mRNAs, including that of Mig2p, a transcription factor linked to glucose sensing and metabolism (12). This observation prompted the suggestion that Rnt1p may act as glucose dependent gene expression regulator.

Glucose dependent gene expression involves one of the most studied networks of transcriptionally regulated genes. In Saccharomyces cerevisiae, glucose induces broad changes in gene expression (13-20) that are primarily triggered by two sensory pathways (Fig. 1A). The first is the Snf3p-Rgt2p pathway, which directly detects glucose levels in the growth medium (17) via two glucose sensors embedded in the cell membrane called Snf3p (21,22) and Rgt2p (23). These sensors generate intracellular signals that permit the expression of glucose transporter genes (Hxt 1-4)(19,24,25). The main target of this signaling pathway is Mth1p (26), a protein that is required for the activation of Rgt1p (27), a transcription factor that binds to the promoter of the Hxt genes and suppress their expression in the absence of glucose. In the presence of glucose, Snf3p and Rgt2p trigger the degradation of Mth1p and thus inactivate Rgt1p, thereby permitting the transcription of the Hxt genes (28,29). 211

The second signaling pathway senses glucose metabolism (19,30) initiated by the phosphorylation and consequent activation of the protein kinase Snf1p (31- 34). In the presence of low glucose concentrations, Snf1p is dephosphorylated and becomes inactive. This allows the transcription repressor Mig1p to accumulate in the nucleus and repress the transcription of glucose metabolism genes such as the sucrose hydrolyzing enzyme Suc2p (35,36). On the other hand, at high glucose concentrations, Snf1p is active and phosphorylates Mig1p. This forces it to exit to the nucleus and to enter the cytoplasm, thus relieving the repression of Suc2p and the other glucose metabolism genes (37). The Snf3p-Rgt2p and the Snf1p dependent pathways are linked by Mig2p (31,37-39). Mig2p is a zinc finger repressor that is closely related to Mig1p and that regulates a distinct subset of Mig1p’s targets in response to higher glucose concentrations (40). Interestingly, Mig2p is not regulated by the Snf1p pathway, but rather by the Snf3p-Rgt2p pathway, and also plays a role in the repression of glucose transporters (40). At low glucose concentrations, Rgt1p represses the transcription of Mig2p and thus triggers the expression of glucose metabolism genes (39).

In both of the glucose sensing pathways described above, gene expression is mainly regulated via transcriptional activity, while RNA decay is considered a passive event that does not directly contribute to the glucose response. Nevertheless, changes in glucose concentration may selectively alter the decay rate of several mRNAs. For example, the degradation of gluconeogenesis mRNAs, such as those coding for Suc2p and Fbp1p, was shown to be accelerated in a glucose dependent manner (41,42). This change in RNA stability appears to be specific to a particular subset of genes since other transcripts, such as the Act1 and Pgk1 mRNAs are not affected (42). The commonly accepted model for glucose dependent RNA decay suggests that depletion of glucose induces a specific translational inhibition that sentences the mRNA for decapping and exonucleolytic degradation (41,43). In this model, RNA decay does not play a direct role in the glucose response, but rather functions as a surveillance mechanism that ensures prompt removal of the no longer needed untranslated 212

RNA. It remains unclear whether or not selective RNA degradation may directly contribute to the signaling cascades of the glucose response.

In order to evaluate the contribution of RNA degradation to glucose sensing, the impact of known yeast ribonucleases including Rnt1p (10), the nuclear exoribonuclease Rrp6p (44) and the cytoplasmic exoribonuclease Xrn1p (45) on the expression of glucose dependent genes was examined. The results indicate that the repression of glucose sensing genes is regulated by targeted RNA degradation, and is largely unaffected by the generic machinery of RNA decay. The Rgt1p glucose sensing regulatory loop that includes both the activator Mth1p and the transcription repressor Mig2p was selectively regulated by Rnt1p. Surprisingly, Rnt1p altered not only the RNA stability, but also influenced the transcription of the Rgt1p associated genes. Indeed, Rnt1p inhibited gene expression in a promoter dependent manner, and directly modulated the transcription independent of the RNA sequence. Together, the results suggest a model in which the promoters of Rgt1p associated genes recruit Rnt1p in order to control glucose dependent steady-state expression and to potentiate a rapid response to any changes in the glucose concentration.

213

Figure 1. Yeast RNase III selectively regulates the transcription factors associated with the Snf3p–Rgt2p glucose-signaling pathway. (A) Schematic representation of the mechanism of glucose-dependent regulation of gene expression involving factors associated with Snf3p– Rgt2p signaling. The transcriptional state of the genes is indicated as either being activated (ON) or repressed (OFF). Arrows and bars indicate activation and inhibition, respectively. The dashed line indicates the constitutive repression of Rgt1 mRNA by Rnt1p. (B) Heat map 214

representing the expression status of the genes involved in glucose signaling and transport in different mutant yeast strains. The increases and decreases in mRNA levels relative to that of the wild-type strain grown under standard laboratory conditions are indicated in magenta and green, respectively. The first column (rnt1∆) indicates mRNAs that change by 1.4- fold upon the deletion of RNT1. The presence or the absence of Rnt1p cleavage sites (Loop), as predicted in silico, are indicated. The cleavage column indicates the capacity of recombinant Rnt1p to cleave the different mRNAs in vitro. Similarly, changes in RNA expression by ≥1.4-fold upon the deletion of RRP6 or XRN1 are also indicated. The genes were organized vertically according to their contributions to glucose signaling as indicated on the left. Cells where left blank when no data were available (ND). (C and D) In vitro cleavage of total RNA extracted from rnt1∆ cells. RNA was extracted from either wild-type cells (RNT1), or cells lacking RNT1 (rnt1∆), and subjected to Northern blot analysis either directly or after incubation with recombinant Rnt1p in vitro (rnt1∆+E). Probes specific either to the Rgt1 (C) or the Mth1 (D) mRNAs where used to detect both the full transcripts and 5’-end cleavage products. Act1 mRNA was used as loading control, and the position of the 25S and 18S ribosomal RNAs are indicated on the left as size markers. P1 and P2 indicate the positions of the cleavage products. The positions of the long (L) and short (S) forms of each gene are indicated on the right. The asterisk represents a non- specific band observed in strands carrying a knockout in RGT1 (data not shown). Primer extension using probes located downstream of the predicted cleavage sites are shown on the right, and the positions of the two cleavage sites (C1 and C2) are indicated on the right. The predicted Rnt1p cleavage signal, the detected cleavage sites and the position of the different probes used are illustrated schematically on top of the gels.

Materials and Methods

Strains and plasmids

Yeast strains were grown and manipulated using standard procedures (46,47). All strains used in this study are listed in Supplementary Table S1. Unless speciﬁed otherwise, all strains were grown at 26°C in YEP media supplemented with either 2% dextrose or 4% galactose, as speciﬁed in each experiment. LacZ- transformed strains were grown in YC(-)ura media. The inactivation of rat1-1 allele was accomplished by growing the cells at 26°C, then shifting them to the restrictive temperature (37°C) for 4 h before harvesting as previously described (48). PACT1- 215

MIG2, PACT1-MTH1 and PACT1-RGT1 strains were created by replacing the respective promoter sequences of MIG2, MTH1 and RGT1 with the ACT1 promoter using standard gene replacement procedures (49). First, a 500 nt PCR fragment corresponding to the ACT1 promoter was amplified using yeast genomic DNA as a template. The PCR product was then inserted downstream of the KanMX gene using the SacI and SpeI restrictions sites in the pCM224 vector (49). The resulting KanMX-ACT1 promoter cassette was further amplified by PCR using probes containing sequence homology with the region located upstream of the target gene (i.e. 400–500 nt upstream of the translation start codon of the MIG2, MTH1 or RGT1). Finally, the resulting PCR fragments were transformed into both wild-type (RNT1) and rnt1∆ strains, and the transformants selected for growth on G418- containing media. Adequate integration of the exogenous promoter was verified by PCR reaction followed by restriction enzyme profile analysis. The pMIG2pr-LacZ, pMTH1pr-LacZ and pRGT1pr-LacZ plasmids were generated as described before (37) by inserting the PCR-amplified promoter regions of MIG2 (500 bp), MTH1 (495 bp) and RGT1 (711 bp) between the BamHI and EcoRI restriction sites of yEP357R vector (50). The resulting plasmids were then transformed into W303 and rnt1∆ strains. All oligonucleotides used in this study are listed in Supplementary Table S2.

In vitro RNA cleavage

Cleavage of total RNA extracted from both wild-type and rnt1∆ cells was conducted essentially as described previously (51). Brieﬂy, 30 mg total RNA was incubated with 4 pmol puriﬁed Rnt1p (48,52) for 20 min at 30°C in 100 ml of reaction buffer [30 mM Tris–HCl (pH 7.5), 5 mM spermidine, 0.1 mM DTT, 0.1 mM

EDTA (pH 7.5), 10 mM MgCl2, 150 mM KCl]. The reactions were stopped by phenol:chloroform extraction, and the RNA collected by salted ethanol precipitation for analysis.

216

RNA analysis

Northern blots were performed as described previously (48) using 15 mg of total RNA and a 1% denaturing agarose gel. The RNA was visualized by autoradiography using randomly labelled probes corresponding to speciﬁc genes (a labelled oligonucleotide probe was used in the case of LacZ). The RNA was quantiﬁed using a Storm 825 scanner (GE Healthcare) and the ImageQuant software (Molecular Dynamics). The primer extension reactions used to map the cleavage sites of Rnt1p in vitro were performed using 5 mg of cleaved total RNA and 1 ng of 32P end-labelled oligonucleotide as described (53). The primers used to generate the probes are listed in Supplementary Table S2.

Chromatin immunoprecipitation

Chromatin extraction and immunoprecipitation were performed as described previously (48). Monoclonal anti-Rpb1 8WG16 (Covance, Berkeley, CA, USA) was used to pull down the RNA polymerase II complex. Quantitative PCR analysis was performed according to the method previously described (54). The radioactivity of each PCR fragment was quantiﬁed using a Storm 825 scanner (GE Healthcare). All signals were normalized using an internal control derived from an unexpressed region of chrV and RNAPII occupancy was calculated by comparing the signals from the immunoprecipitated samples relative to that of the input samples for each primer pair.

Results

Rnt1p selectively degrades the mRNAs associated with the Snf3p–Rgt2p glucose sensing pathway

It has previously been shown that Rnt1p degrades the mRNA encoding the glucose-dependent transcription factor Mig2p (12). This suggests that post- transcriptional gene regulation may play an important role in the glucose response 217

pathway. In order to evaluate this hypothesis we identified all of the genes associated with glucose signaling (29), glucose response and transport (55) and their expression patterns determined in the absence of different yeast ribonucleases. Previously generated genome-wide expression profiles (12) of both wild-type cells and cells carrying deletions of the three main non-essential ribonucleases in yeast, namely RNT1, RRP6 and XRN1, were used to identify potential targets for RNA degradation. As shown in Figure 1B, the deletion of RNT1 gene increased the expression of ~20% of the genes associated with glucose response, while deletion of either the nuclear 3’–5’ exoribonuclease RRP6 or the 5’–3’ cytoplasmic exoribonuclease XRN1 increased the expression of only one or two of these genes. This suggests that Rnt1p is preferentially implicated in regulating the expression of the glucose-associated genes, and that the expression of these genes is not highly dependent on exonucleolytic RNA degradation. In general, the genes up-regulated by RNT1 deletion where comparably distributed across the different classes of glucose-dependent genes. In order to identify direct targets of Rnt1p, all genes associated with the glucose response were examined for the presence of NGNN stem loop structures (11,48), which constitute Rnt1p cleavage signals. As shown in Figure 1B, mRNAs of all glucose-associated genes, with the exception of six genes, exhibited local structures (loop) that may be recognized by Rnt1p. However, in vitro cleavage assay using recombinant Rnt1p indicated that only three RNAs are direct substrates of Rnt1p. This result was not unexpected as the majority of local stem loops do not fold in this context and thus cannot support cleavage by the recombinant enzyme (48,56). Two of the RNA substrates that were cleaved by Rnt1p encode Mth1p (26) and Rgt1p (27), transcription factors associated with the Snf3p–Rgt2p glucose induction pathway (Figure 1B). The third encodes Mig2p (57), which has previously been shown to be regulated by both Rgt1p (39) and Rnt1p (12). These data indicate that Rnt1p does not generically influence the RNA stability of the glucose-dependent genes, but instead selectively targets a tightly linked glucose sensing regulatory loop. 218

In order to confirm the impact of Rnt1p on the expression and cleavage of the two newly identified substrates, the impact of Rnt1p on both Mth1 and Rgt1 mRNA in vivo and in vitro was examined using northern blot analysis. As shown in Figure 1C, Rgt1 expression was detected in wild-type (RNT1) cells in two forms corresponding to long (Rgt1-L) and short RNA (Rgt1-S) transcripts. Based on previous tiling array expression profiles (58), Rgt1-L is likely a 3’-extended polycistronic transcript arising from transcription termination after the downstream gene (AIM26). As expected, both forms increased in rnt1∆ cells. Reverse transcription using a primer complementary to the sequence downstream of the predicted loop confirmed the position of the cleavage site and ensured the specificity of the cleavage reaction. In the case of MTH1 (Figure 1D), which is not expressed in cells grown on standard media containing glucose, the mRNA was only detected in rnt1∆ cells, clearly indicating that Rnt1p is required for the glucose-dependent shut down of MTH1. Similar to Rgt1, two transcripts (Mth1-S and Mth1-L) were detected, and further investigation confirmed that the longer form is a polycistronic transcript consisting of MTH1 and the downstream PMP3 gene (Figure 5 and data not shown). Once again, Northern blot and primer extension analysis of RNA incubated in the presence of recombinant Rnt1p confirmed cleavage at the predicted site. Clearly Rnt1p directly regulates the expression of both RGT1 and MTH1 genes, at least in part, by endonucleolytic cleavage of their messenger RNAs.

Rnt1p promotes glucose-dependent repression of Rgt1p-associated factors

In order to determine the impact of Rnt1p on glucose-dependent gene expression, RNT1 and rnt1∆ cells where grown in media containing either glucose or galactose and the expression levels of Mig2, Rgt1 and Mth1 mRNAs were analysed by northern blot. As reported earlier (12), the expression of MIG2 in RNT1 cells was detected when the cells were grown in the presence of glucose (ON condition), and the expression was inhibited when the cells were grown in the presence of galactose (OFF condition) (Figure 2A). The deletion of RNT1 (rnt1∆) 219

increased the expression of MIG2 in both the ON and OFF conditions to a similar extent (Figure 2A, right panel). The glucose-mediated induction was found to be about three fold in both RNT1 and rnt1∆ cells (Figure 2A, left panel), suggesting that Rnt1p is equally required in both conditions. In the case of RGT1 (Figure 2B), expression was detected in the presence of both glucose (OFF condition) and galactose (ON condition), as expected, as RGT1 is known to be regulated at the protein level (59). The deletion of RNT1 increased the expression of the long form of Rgt1 (Rgt1-L) in both growth conditions without affecting the expression of the short form (Rgt1-S). Thus, Rnt1p regulates both the quantity and pattern of RGT1 expression in a glucose independent manner. In contrast, MTH1 expression was detected only in the presence of galactose (ON condition), and the effect of RNT1 deletion was found to be more pronounced when the gene was OFF (Figure 2C). This indicates that unlike RGT1, Rnt1p plays an important role in regulating the glucose-dependent repression of MTH1. Together, these results indicate that Rnt1p plays different roles in regulating gene expression rates that vary from constitutive (e.g. MIG2) to condition enhanced inhibition (e.g. MTH1) of gene expression.

220

Figure 2. Rnt1p optimizes the expression of the Rgt1-associated transcription factors. Total RNA was extracted from either RNT1 or rnt1∆ cells grown in the presence of either glucose or galactose and the expression levels of the Mig2 (A), Rgt1 (B) or Mth1 (C) mRNAs were detected by Northern blot (left panels). Act1 mRNA was used as a loading control. The bands were quantiﬁed, and the relative RNA amount of three biological replicas was calculated. Bar graphs (middle panels) illustrate the impact of RNT1 deletion on the expression of each gene in the presence of different sugars relative to that of wild-type cells. Bar graphs on the right illustrate the sugar-dependent fold induction (i.e. ratio of mRNA amount detected in ‘ON’ condition over ‘OFF’ condition) of each mRNA in both RNT1 and rnt1∆ cells. 221

Rnt1p mediates the promoter-dependent repression of gene expression

Glucose-dependent genes are primarily regulated at the transcriptional level by promoter-speciﬁc transcription factors (60,61). For this reason the impact of the promoter sequence on the RNT1-dependent expression of Mig2, Rgt1 and Mth1 mRNAs was tested. Each gene’s promoter was replaced by that of the house keeping gene ACT1 (62), and the expression was monitored using total RNA extracted from both RNT1 and rnt1∆ cells grown in different sugar conditions. As expected, the expression of Mig2 mRNA driven from ACT1 promoter abolished most of the glucose-dependent response (compare Figures 3A and 2A), demonstrating that the endogenous promoter is essential for conditional repression. The same trend was also observed with PACT1-MTH1 where promoter replacement also abolished the glucose-dependent repression (compare Figures 3C and 2C). In the case of RGT1, whose expression is not regulated by glucose, the promoter replacement increased the relative expression level in both sugar conditions (compare Figures 3B and 2B). The deletion of RNT1 increased the expression levels of all three genes, even when they were expressed from exogenous promoters, regardless of the sugar conditions (Figure 3, right panels). Consistently, mutations that alter Rnt1p cleavage signal increased the Mth1 mRNA half-life (Supplementary Figure S1B) (12). This conﬁrms that at least part of the Rnt1p inhibition of gene expression is dependent on the sequence harboring Rnt1p cleavage site in good agreement with Rnt1p targeted RNA degradation. Indeed, the deletion of RNT1 increases the half-life of both the Mig2 (12) and the Mth1 mRNAs (Supplementary Figure S1A). Surprisingly, the increase in the expression levels upon RNT1 deletion was more pronounced in genes expressed from their endogenous promoters, suggesting that the promoter enhances Rnt1p-dependent repression. Interestingly, in the case of MTH1, the impact of the promoter on the RNT1-mediated repression was only observed under the OFF condition, suggesting that Rnt1p inhibits the accumulation of Mth1 mRNA in a glucose- dependent manner in vivo. We propose that Rgt1p-associated factors are 222

regulated via a coordinated mechanism of gene repression that combines both transcriptional and post-transcriptional levels of gene regulation.

Figure 3. Promoters partially mediate the Rnt1p-dependent repression of gene expression. The expression of MIG2 (A), RGT1 (B) or MTH1 (C) driven by either the endogenous or the ACT1 promoters where assayed by Northern blots in both RNT1 and rnt1∆ strains grown in media containing either glucose or galactose (left panels). The positions of the 25S and 18S ribosomal RNAs are indicated on the left, and the loading control is shown at the bottom. The averages mRNA levels of the different genes in each sugar condition were quantiﬁed and plotted in order to illustrate the effect of RNT1 deletion relative to wild-type cells (right panels). The data from Figure 2 (endogenous promoters) was repeated so as to facilitate comparison. The data shown are the average of three independent experiments. ON and OFF indicate the conditions that are either inducing or repressing the expression of each gene, respectively.

223

Rnt1p mediates cleavage signal independent transcription inhibition

In order to directly examine the contributions of the MIG2, RGT1 and MTH1 promoters to both glucose and Rnt1p-dependent repression, the endogenous coding sequence starting from the translation start codon (AUG) was replaced with that of a reporter gene (LacZ) and the promoters’ activities were monitored under different conditions. Replacement (pMIG2pr-LacZ) of the coding sequence reduced the MIG2 response to glucose to 1.4-fold instead of 3-fold (Figures 4A and 2A) when Rnt1p was present, suggesting that RNA degradation plays an important role the regulation of this gene. Therefore, the presence of the MIG2 promoter is necessary (Figure 3A, left panel, lanes 1 and 2), but not sufﬁcient (Figure 4A, left panel, lanes 1 and 2) for optimal glucose response. Surprisingly, the deletion of RNT1 increased the expression of the MIG2 promoter (Figure 4A, left panel, lanes 3 and 4) in the absence of the RNA cleavage site detected in vitro. When driven from MIG2 promoter (pMIG2pr-LacZ), expression of LacZ mRNA in rnt1∆ was more pronounced in the ON condition than that in the OFF condition, suggesting that Rnt1p induces the basal promoter’s activity and was not simply alleviate repression (Figure 4, right panel). This increase is not due to a global increase either in the promoter activities or gene expression since the majority of genes are under transcribed in rnt1∆ (48) and the expression of the LacZ reporter gene did not increase when driven by unrelated promoter like ACT1 (data not shown) (63). It should also be noted that Rnt1p effect is unlikely to be caused by transcription independent activity of the 5’-UTR of the transcripts since the enzyme did not cleave this region (Figure 1). However, we cannot exclude the possibility that the 5’-UTR play a role in mediating Rnt1p impact on transcription. In the case of pRGT1pr-LacZ, expression on LacZ mRNA was moderately increased in rnt1∆ cells grown in OFF condition, and was strongly increased in ON conditions when compared with RNT1 cells grown under these conditions (Figure 4B). This result is unexpected since the wild-type allele of RGT1 does not seem to respond to glucose at either the transcriptional or post-transcriptional levels [Figure 2 and (64)]. One explanation for this apparent contradiction is that Rgt1 mRNA 224

degradation conceals the effect on the promoter activity observed in absence of Rnt1p. Indeed, Rgt1 expression from a heterologous promoter responded equally to RNT1 deletion under both sugar conditions (Figure 3B). Unlike MIG2 and RGT1, Rnt1p does not inhibit the promoter activity of MTH1 in the ON condition, but rather specifically reduces the promoter repression under the OFF conditions (Figure 4). This result is consistent with a role for Rnt1p in regulating the glucose-dependent expression of Mth1 mRNA observed in Figure 2. The conclusion drawn is that yeast dsRNA-specific ribonuclease may influence gene expression in two non- exclusive manners: one is promoter dependent and cleavage site independent, while the other requires the original open reading frame sequence.

Figure 4. Rnt1p mediates cleavage site independent repression of gene expression. RNA was extracted from both RNT1 and rnt1∆ strains expressing LacZ RNA from the MIG2 (A), RGT1 (B) or MTH1 (C) promoters. The cells were grown in different media containing either glucose or galactose in order to assay the sugar effect. The relative amounts of RNA were detected by northern blot analysis using probes complimentary to LacZ (left panel), and are represented in a bar graph (right panel) as described in Figure 2. The data shown represent an average of at least three independent experiments. 225

Rnt1p inhibits RNAPII association with MTH1 DNA in a glucose-dependent manner

Since MTH1 is the only gene regulated by Rnt1p at the promoter level in a glucose-dependent manner, whether or not this regulation is directly related to an increase in transcriptional activity, and whether or not glucose regulates the Rnt1p contribution to transcription repression was investigated. Accordingly, the occupancy of the RNA polymerase II complex (RNAP II) along MTH1 locus in both RNT1 and rnt1∆ cells grown in different conditions was monitored and directly compared to the corresponding transcripts accumulation. RNAP II association with the transcription unit was examined by chromatin immunoprecipitation (ChIP) using antibodies directed against the Rpb1p subunit (65). The precipitated DNA fragments where amplified using probes covering the complete MTH1 gene, the adjacent intergenic area and the downstream gene PMP3 so as to clearly delineate the transcription unit (Figure 5A). As shown in Figure 5B, under ON condition, RNAPII co-immunoprecipitated, in both RNT1 and rnt1∆ strains, DNA fragments corresponding to the promoter region, the coding sequence and the intergenic region downstream of MTH1 (fragments B–H), but not the untranscribed region upstream of the promoter (fragment A). The fragments corresponding to the genes located downstream were equally associated with RNAPII suggesting that, under ON conditions, the expression of MTH1 does not terminate efﬁciently and reaches levels similar to that of the downstream genes. The transcription read-through by RNAPII on MTH1 was conﬁrmed by the accumulation of a long RNA transcript corresponding to a 3’-end extension (Figures 1 and 5C; data not shown). Under OFF conditions, few DNA fragments corresponding to the MTH1 sequence where immunoprecipitated with RNAPII in RNT1 cells. In contrast, those corresponding to PMP3 precipitated under OFF conditions to levels similar to that observed for the ON condition, as would be expected from a glucose independent gene (Figure 5B). The differences between RNAPII association in the presence and the absence of Rnt1p under the OFF condition are statistically significant with a combined P-value of 0.0025. This confirms that glucose does indeed repress MTH1 transcription. 226

Interestingly, the deletion of RNT1 specifically increased the RNAPII association with MTH1 under the OFF condition, but not with PMP3, clearly showing that Rnt1p selectively inhibits transcription in a sugar-dependent manner.

The RNAPII association pattern with MTH1 indicates that transcription termination and sugar-dependent repression of this gene are partly dependent on RNA degradation. Consequently, the impact of the main yeast ribonucleases implicated in both the nuclear and the cytoplasmic degradation of RNA on the expression of MTH1 locus was investigated. The different deletion strains where grown under OFF condition in order to determine the factors that contribute to MTH1 repression. As shown in Figure 5C, very little RNA was detected in wild-type cells (WT, lane 1), while the two forms of Mth1 (L and S) where detected in rnt1∆ cells and were cleaved by the recombinant enzyme (Lanes 2 and 3). As expected, Mth1-L in rnt1∆ cells was also weakly detected with probes downstream of MTH1 and within PMP3. Interestingly, the deletion of the 5’–3’ cytoplasmic exoribonuclease XRN1 resulted in the accumulation of Mth1-L mRNA (Lane 4). This clearly indicates that, even under OFF conditions, a certain level of Mth1 mRNA is constitutively produced and degraded in the cytoplasm. The double deletion of RNT1 and XRN1 increases the amount of Mth1-L (Lanes 5, 18, 31), once again conﬁrming that Mth1-L is regulated by Rnt1p in the nucleus and Xrn1p in the cytoplasm. The deletion of the nuclear or the cytoplasmic 3’–5’ ribonucleases RRP6 and SKI7 did not have much effect on expression, suggesting that 3’-end degradation does not play an important role in repressing the expression of MTH1. In contrast, cells carrying a temperature sensitive allele of the nuclear 5’–3’ exoribonuclease RAT1 (rat1-1) displayed a modest increase in the amount Mth1-S (Lanes 8 and 9). However, a signiﬁcant increase in both forms of Mth1 and in RNA transcripts corresponding to 5’-end extended Pmp3 were detected at both the permissive and the restrictive temperatures in strains carrying both a deletion in XRN1 and the rat1-1 allele (Lanes 10, 11, 23, 24, 36 and 37). The results clearly demonstrate that post-transcriptional regulation may play a much more important role in gene expression than previously anticipated. We propose that 227

transcriptional and post-transcriptional regulation works as a tightly integrated unit in order to achieve a rapid and complete repression of gene expression.

Figure 5. Rnt1p enhances the glucose-dependent transcriptional repression of MTH1. (A) Schematic representation of the MTH1 gene locus. The two forms of Mth1 mRNAs detected by Northern blot (C) are illustrated on top. The position of the Rnt1p cleavage signal (tetraloop) and the predicted polyadenylation signals (pA), are shown. The positions of the probes used for the Northern blots shown in (C), and the regions ampliﬁed after the ChIP shown in (B) are indicated at the bottom. (B) ChIP was performed using antibodies against the RNAP II protein subunit Rpb1p in either RNT1 or rnt1∆ cells grown in media containing either 2% dextrose (2% dex) or 4% galactose (4% gal). The precipitated DNA was ampliﬁed by quantitative radiolabelled PCR using the primers indicated in (A), and the average values of three independent biological replicates were used to calculate the enrichment relative to the input samples. A primer pair amplifying a known untranscribed region of chromosome V (chrV) was used to normalize the signals. (C) Northern blot analysis of total RNA extracted from strains either lacking or carrying mutations in different ribonucleases. The positions of the probes are indicated in (A). 228

Schematics of the different RNA transcripts observed are indicated on the right. Both Act1 mRNA and 25S ribosomal RNA were used as loading controls. The extended transcripts observed in the rat1-1/xrn1∆ RNA represent the transcriptional read-through expected upon the inactivation of rat1-1 (84).

Discussion

This study demonstrates that targeted RNA degradation plays an important role in enhancing conditional transcription repression of glucose-dependent genes. The dsRNA-speciﬁc ribonuclease III Rnt1p selectively repressed the expression of factors associated with the Snf3p–Rgt2p sensing pathway in vivo, and directly cleaved the associated mRNAs in vitro (Figure 1). In contrast, the deletion of ribonucleases like XRN1 or RRP6, which are required for general RNA turnover, did not significantly alter the repression of the glucose-associated genes, underscoring the preference for the Rnt1p contribution within the glucose regulatory network. The Rnt1p-mediated repression of gene expression was partially dependent on the promoter sequence, suggesting that Rnt1p is recruited to its substrate during transcription (Figure 3). Strikingly, the promoter activity was independently suppressed by Rnt1p expression independent of the RNA sequence (Figure 4). Indeed, the association of RNAPII with the MTH1 DNA increased in the absence of RNT1, confirming that Rnt1p does not only decrease gene expression by sentencing RNA for degradation, but may also repress transcription. The glucose-dependent expression pattern of the Rnt1p substrates indicates that the enzyme contributes to glucose response in a gene-specific manner that varies from the fail-safe repression of transcription (MIG2 and RGT1) to direct glucose- dependent repression (MTH1) (Figures 2 and 6). Taken, together the results presented here reveal a new mode of gene regulation in which RNA degradation factors may simultaneously degrade nascent RNA transcripts and inhibit de novo transcription.

The regulation of the glucose response was mostly thought to be carried out by a well knit transcriptional network, with a few exceptions in which either protein 229

or RNA degradation were considered to be factors in the signaling pathway (66,67). Several examples of differential RNA degradation were noted in the gluconeogenic pathway, including the Fbp1 and Pck1 mRNAs that are specifically degraded at low levels of glucose (42). The mRNAs of other genes that are not directly connected to glucose metabolism, like the iron protein subunit gene SDH1, were also shown to degrade in response to glucose. However, in this case, the degradation was accelerated only in the presence of high glucose levels (41). In all cases, the signal that trigger the accelerated degradation was not identified, nor was the ribonuclease identified, with the exception of the cytoplasmic 5’–3’ exoribonuclease Xrn1p that was linked to the degradation of the Sdh1 mRNA (41). Similarly, the glucose-sensing pathway was considered to be solely regulated by transcriptional activity. For example, a recent model suggested that the glucose transporter genes are differentially regulated by a transcriptional pulse of the transcription repressors Rgt1p and Mig2p in response to the amount of glucose present in the cells (67). In this mathematical model, one that considers RNA degradation as being constant, the efficiency with which Rgt1p and Mig2p repress the expression of each HXT gene determines which target genes have a pulse of transcription in response to glucose (67). In contrast, this study demonstrates that RGT1 and its activator MTH1, as well as MIG2, gene expression is determined in large part by selective RNA degradation. This clearly changes the current view of how glucose sensing is achieved. As described in the model illustrated in Figure 6, RNA degradation may contribute to glucose sensing either by providing a means for fast repression, by constant surveillance, or by the conditional repression of the relevant genes.

230

Figure 6. Proposed models of Rnt1p-dependent gene regulation. The Rnt1p contribution to the RNA degradation and transcriptional repression of MIG2, RGT1 and MTH1 genes are illustrated in both ON (genes are expressed) and OFF (expression inhibited) conditions. (A) Rnt1p may mediate a rapid and robust repression by down-regulating transcription and RNA stability in the ON condition, and by enhancing the effect of glucose-dependent transcription repressors (R) by degrading all nascent transcripts present in the nucleus in the OFF condition. (B) Constitutive down-regulation and surveillance of gene expression may be achieved by constant transcriptional repression and RNA degradation that prevents the production of aberrant RNA transcripts and maintains a constant supply of proteins for genes whose activities are regulated at the proteins level. (C) Rnt1p may differentially inhibit transcription and induce RNA degradation in a condition-speciﬁc manner. In this case, Rnt1p inhibits the transcription only under the OFF condition, and its impact on RNA degradation is more robust when the genes are turned OFF.

In the fast repression mode (Figure 6A), as in the case of MIG2 gene, Rnt1p decreases the steady-state mRNA level of Mig2 mRNA by cleaving a percentage of the newly synthesized RNA co-transcriptionally and thereby reducing the 231

transcription rate. In this case, the activity of Rnt1p appears to be constitutive, and independent of glucose, as the deletion of RNT1 increased the expression of Mig2 mRNA to the same extent in both ON and OFF conditions (Figure 2). This mode of constant promoter coupled RNA degradation allows all transcripts to be rapidly degraded once transcription is halted by repressors like Rgt1p (39,67). In addition, Rnt1p was shown to be required for the fast degradation of the Mig2 mRNA immediately after a transition from a glucose to a glycerol containing media (12). Thus, RNA degradation ensures the fast and sustained repression of conditionally regulated genes. This mode of repression is particularly required for glucose sensors due to the constant flux of glucose that cells normally experience in their natural habitat. Indeed, the short bursts of MIG2 transcription hypothesized by the incoherent feed forward regulatory loop model (67) are difficult to envision if all nascent transcripts (i.e. transcripts that are still produced before the transcriptional repression is activated) have to be degraded post-transcriptionally in the cytoplasm as suggested for glucose sensitive genes like SDH1 (41). Moreover, a recent study in mammalian cells demonstrated that RNA degradation is required to sharpen the transcription peak, further supporting the hypothesis that the coordination of transcriptional repression and RNA degradation is essential for producing optimal non-overlapping transcriptional pulses (68).

Rnt1p also contributed to glucose sensing by the constant surveillance of RNA transcripts that are not conditionally repressed by glucose (Figure 6B). In this mode, represented by RGT1, Rnt1p constantly cleaves any excess RNA co- transcriptionally thereby preventing it from being translated. In this way, a constant amount of Rgt1p is produced allowing for a sensitive activation through the protein- protein interaction with Mth1p. The need for this method of transcriptional repression is not to increase the rate of the transcriptional repression cycle, but rather to balance the production of Rgt1p with that of its activator Mth1p. Mth1 RNA is also cleaved by Rnt1p, and the cleavage in this case appears to provide a means for the glucose-dependent conditional repression of the MTH1 gene (Figure 6C). Unlike for the Mig2 and Rgt1 mRNAs, Rnt1p cleaves only a small fraction of 232

the Mth1 mRNA under ON conditions without interfering with transcription. However, once the cells are moved to the OFF conditions, Rnt1p appears to specifically repress the transcription of MTH1 and cleaves its RNA in a glucose- dependent manner. This is supported by the fact that the deletion of Rnt1p had a greater effect on the repression of Mth1 mRNA in OFF condition than in the ON condition (Figures 2E and 4E). Interestingly, it was demonstrated that protein degradation by itself is not sufficient to explain the reduction in Mth1p observed after glucose addition. Moreover, even when protein decay and transcriptional repression are combined, the predicted rate of Mth1p depletion remains relatively slow (69). We propose that the conditional regulation of the Mth1 mRNA level by Rnt1p prevents any residual mRNA from escaping the nucleus, thus allowing for a faster repression of Mth1p expression. Overall, through these three different modes of gene repression, Rnt1p provides the glucose sensing network the means to ﬁne tune transcription as mandated by the glucose availability and fluctuation.

Traditionally, eukaryotic RNA degradation was considered as an independent post-transcriptional step that takes place once transcription is complete (70). This view was fuelled by the image of RNA degradation being mostly cytoplasmic, while transcription occurs in the nucleus (71–73). However, it has become increasingly clear in recent years that certain RNA actively degrades in the nucleus (74–76), and that this degradation is not restricted to erroneous or misfolded RNA as previously thought (51,77). The degradation of RNA in the nucleus makes the distinction between transcriptional and post-transcriptional events much more difficult. It is now established that RNAPII interacts via its C- terminal domain (CTD) with the RNA modification and processing factors that are required for the maturation of mRNA (78,79). This commits the nascent transcript very early to maturation and cytoplasmic export (80), which makes the RNA degradation of mRNA difficult to achieve unless the RNA is either deliberately retained in the nucleus, or the involved ribonucleases are recruited to the transcription unit. In the cases of MTH1, RGT1 and MIG2 genes, we propose the latter scenario where Rnt1p is actively recruited to the transcription site. It was 233

previously shown that Rnt1p associates with the chromatin of actively transcribed genes in order to promote their polyadenylation independent transcription termination (48). In parallel, the promoters of the MTH1, RGT1 and MIG2 genes seem to play an important role in enhancing the Rnt1p-dependent repression, and, as such, suggest that Rnt1p is linked to the transcriptional activity. In addition, ChIP-on-CHIP assays suggest that Rnt1p is recruited to the DNA of many genes (48), including MIG2 (data not shown). This recruitment to the transcription site also permits Rnt1p to directly influence transcription as was noted in the cases of MTH1 (Figure 5) and other genes (48). The sequence elements required for Rnt1p- dependent transcription repression appears to be embedded in the core promoter since deletion analysis failed to separate Rnt1p repression from basic transcription (data not shown). Other ribonucleases like Rrp6p, Xrn1p and Rat1p were shown to affect transcription by silencing bidirectional promoters and triggering transcription termination (81–84). However, in all cases, these activities were associated with the degradational activity of these enzymes. Conversely, in the case of Rnt1p, transcription is altered even in the absence of its RNA cleavage site (Figure 4). It is unlikely that the effect of Rnt1p on transcription is generic or indirect due to a general perturbation of transcription since RNT1 deletion only increases the transcription of a minority of genes and most of these are related to Rnt1p substrates (12). In fact, very few genes display differential increase in transcription when RNT1 is deleted (48). It is possible, however, that Rnt1p conditionally associates with the RNAP II complex and thus triggers conformational changes in the transcriptional machinery leading to changes in transcription pattern. Alternatively, Rnt1p may function as genuine transcription repressor independent of RNA cleavage. There is no direct evidence for this possibility, but this may explain why the enzyme does not directly cleave, in vitro, a large number of genes that are up-regulated upon the deletion of RNT1 (Figure 1 and data not shown). In all cases, the data reported here cement Rnt1p as an integral part of the transcription repression machinery that blurs the borders between the transcriptional and the post-transcriptional regulation of gene expression. 234

Supplementary data

Supplementary Data are available at NAR Online.

Funding

Canadian Institute of Health Research; salary fellowship from the Fonds de la Recherche en Sante´ du Québec (to S.A.E.); Natural Sciences and Engineering Research Council of Canada (graduate student fellowships to both ML and D.G.). Funding for open access charge: Canadian Institute of Health Research.

Conﬂict of interest statement. None declared.

235

References

1. Ji,X. (2008) The mechanism of RNase III action: how dicer dices. Curr. Top. Microbiol. Immunol., 320, 99–116.

2. Banerjee,D. and Slack,F. (2002) Control of developmental timing by small temporal RNAs: a paradigm for RNA-mediated regulation of gene expression. Bioessays, 24, 119–129.

3. Saunders,L.R. and Barber,G.N. (2003) The dsRNA binding protein family: critical roles, diverse cellular functions. FASEB J., 17, 961–983.

4. Agrawal,N., Dasaradhi,P.V., Mohmmed,A., Malhotra,P., Bhatnagar,R.K. and Mukherjee,S.K. (2003) RNA interference: biology, mechanism, and applications. Microbiol. Mol. Biol. Rev., 67, 657–685.

5. Denli,A.M. and Hannon,G.J. (2003) RNAi: an ever-growing puzzle. Trends Biochem. Sci., 28, 196–201.

6. Holen,T. and Mobbs,C.V. (2004) Lobotomy of genes: use of RNA interference in neuroscience. Neuroscience, 126, 1–7.

7. Martienssen,R.A., Zaratiegui,M. and Goto,D.B. (2005) RNA interference and heterochromatin in the ﬁssion yeast Schizosaccharomyces pombe. Trends Genet., 21, 450–456.

8. Petrie,V.J., Wuitschick,J.D., Givens,C.D., Kosinski,A.M. and Partridge,J.F. (2005) RNA interference (RNAi)-dependent and RNAi-independent association of the Chp1 chromodomain protein with distinct heterochromatic loci in ﬁssion yeast. Mol. Cell. Biol., 25, 2331–2346.

9. Hansen,K.R., Burns,G., Mata,J., Volpe,T.A., Martienssen,R.A., Bahler,J. and Thon,G. (2005) Global effects on gene expression in ﬁssion yeast by silencing and RNA interference machineries. Mol. Cell. Biol., 25, 590–601.

10. Lamontagne,B., Larose,S., Boulanger,J. and Elela,S.A. (2001) The RNase III family: a conserved structure and expanding functions in eukaryotic dsRNA metabolism. Curr. Issues Mol. Biol., 3, 71–78.

12. Ge,D., Lamontagne,B. and Abou Elela,S. (2005) RNase III-mediated silencing of a glucose-dependent repressor in yeast. Curr. Biol., 15, 140–145.

236

13. Johnston,M. (1999) Feasting, fasting and fermenting. Glucose sensing in yeast and other cells. Trends Genet., 15, 29–33.

14. Rolland,F., Winderickx,J. and Thevelein,J.M. (2001) Glucose-sensing mechanisms in eukaryotic cells. Trends Biochem. Sci., 26, 310–317.

15. Holsbeeks,I., Lagatie,O., Van Nuland,A., Van de Velde,S. and Thevelein,J.M. (2004) The eukaryotic plasma membrane as a nutrient-sensing device. Trends Biochem. Sci., 29, 556–564.

16. Gelade,R., Van de Velde,S., Van Dijck,P. and Thevelein,J.M. (2003) Multi-level response of the yeast genome to glucose. Genome Biol., 4, 233.

17. Rolland,F., Winderickx,J. and Thevelein,J.M. (2002) Glucose-sensing and - signaling mechanisms in yeast. FEMS Yeast Res., 2, 183–201.

18. Forsberg,H. and Ljungdahl,P.O. (2001) Sensors of extracellular nutrients in Saccharomyces cerevisiae. Curr. Genet., 40, 91–109.

19. Ozcan,S. and Johnston,M. (1999) Function and regulation of yeast hexose transporters. Microbiol. Mol. Biol. Rev., 63, 554–569.

20. Thorens,B. (2004) Mechanisms of glucose sensing and multiplicity of glucose sensors. Ann. Endocrinol., 65, 9–12.

21. Coons,D.M., Vagnoli,P. and Bisson,L.F. (1997) The C-terminal domain of Snf3p is sufﬁcient to complement the growth defect of snf3 null mutations in Saccharomyces cerevisiae: SNF3 functions in glucose recognition. Yeast, 13, 9–20.

22. Marshall-Carlson,L., Celenza,J.L., Laurent,B.C. and Carlson,M. (1990) Mutational analysis of the SNF3 glucose transporter of Saccharomyces cerevisiae. Mol. Cell. Biol., 10, 1105–1115.

23. Polish,J.A., Kim,J.H. and Johnston,M. (2005) How the Rgt1 transcription factor of Saccharomyces cerevisiae is regulated by glucose. Genetics, 169, 583–594.

24. Jansen,M.L., De Winde,J.H. and Pronk,J.T. (2002) Hxt-carrier-mediated glucose efﬂux upon exposure of Saccharomyces cerevisiae to excess maltose. Appl. Environ. Microbiol., 68, 4259–4265.

25. Ozcan,S. and Johnston,M. (1995) Three different regulatory mechanisms enable yeast hexose transporter (HXT) genes to be induced by different levels of glucose. Mol. Cell. Biol., 15, 1564–1572.

26. Lafuente,M.J., Gancedo,C., Jauniaux,J.C. and Gancedo,J.M. (2000) Mth1 receives the signal given by the glucose sensors Snf3 and Rgt2 in Saccharomyces cerevisiae. Mol. Microbiol., 35, 161–172. 237

27. Lakshmanan,J., Mosley,A.L. and Ozcan,S. (2003) Repression of transcription by Rgt1 in the absence of glucose requires Std1 and Mth1. Curr. Genet., 44, 19–25.

28. Johnston,M. and Kim,J.H. (2005) Glucose as a hormone: receptor-mediated glucose sensing in the yeast Saccharomyces cerevisiae. Biochem. Soc. Trans., 33, 247–252.

29. Zaman,S., Lippman,S.I., Zhao,X. and Broach,J.R. (2008) How Saccharomyces responds to nutrients. Annu. Rev. Genet., 42, 27–81.

30. Hardie,D.G. (1999) Roles of the AMP-activated/SNF1 protein kinase family in the response to cellular stress. Biochem. Soc. Symp., 64, 13–27.

31. Lutﬁyya,L.L. and Johnston,M. (1996) Two zinc-ﬁnger-containing repressors are responsible for glucose repression of SUC2 expression. Mol. Cell. Biol., 16, 4790–4797.

32. Smith,F.C., Davies,S.P., Wilson,W.A., Carling,D. and Hardie,D.G. (1999) The SNF1 kinase complex from Saccharomyces cerevisiae phosphorylates the transcriptional repressor protein Mig1p in vitro at four sites within or near regulatory domain 1. FEBS Lett., 453, 219–223.

33. Frolova,E., Johnston,M. and Majors,J. (1999) Binding of the glucose- dependent Mig1p repressor to the GAL1 and GAL4 promoters in vivo: regulation by glucose and chromatin structure. Nucleic Acids Res., 27, 1350– 1358.

34. Ostling,J. and Ronne,H. (1998) Negative control of the Mig1p repressor by Snf1p-dependent phosphorylation in the absence of glucose. Eur. J. Biochem., 252, 162–168.

35. Ozcan,S., Vallier,L.G., Flick,J.S., Carlson,M. and Johnston,M. (1997) Expression of the SUC2 gene of Saccharomyces cerevisiae is induced by low levels of glucose. Yeast, 13, 127–137.

36. Neigeborn,L. and Carlson,M. (1984) Genes affecting the regulation of SUC2 gene expression by glucose repression in Saccharomyces cerevisiae. Genetics, 108, 845–858.

37. Lutﬁyya,L.L., Iyer,V.R., DeRisi,J., DeVit,M.J., Brown,P.O. and Johnston,M. (1998) Characterization of three related glucose repressors and genes they regulate in Saccharomyces cerevisiae. Genetics, 150, 1377–1391.

38. Wu,J. and Trumbly,R.J. (1998) Multiple regulatory proteins mediate repression and activation by interaction with the yeast Mig1 binding site. Yeast, 14, 985– 1000. 238

39. Kaniak,A., Xue,Z., Macool,D., Kim,J.H. and Johnston,M. (2004) Regulatory network connecting two glucose signal transduction pathways in Saccharomyces cerevisiae. Eukaryot. Cell, 3, 221–231.

40. Westholm,J.O., Nordberg,N., Muren,E., Ameur,A., Komorowski,J. and Ronne,H. (2008) Combinatorial control of gene expression by the three yeast repressors Mig1, Mig2 and Mig3. BMC Genomics, 9, 601.

41. de la Cruz,B.J., Prieto,S. and Schefﬂer,I.E. (2002) The role of the 5’ untranslated region (UTR) in glucose-dependent mRNA decay. Yeast, 19, 887–902.

42. Yin,Z., Hatton,L. and Brown,A.J. (2000) Differential post-transcriptional regulation of yeast mRNAs in response to high and low glucose concentrations. Mol. Microbiol., 35, 553–565.

43. Lui,J., Campbell,S.G. and Ashe,M.P. (2010) Inhibition of translation initiation following glucose depletion in yeast facilitates a rationalization of mRNA content. Biochem. Soc. Trans., 38, 1131–1136.

44. Allmang,C., Kufel,J., Chanfreau,G., Mitchell,P., Petfalski,E. and Tollervey,D. (1999) Functions of the exosome in rRNA, snoRNA and snRNA synthesis. EMBO J., 18, 5399–5410.

45. Long,R.M. and McNally,M.T. (2003) mRNA decay: x (XRN1) marks the spot. Mol. Cell, 11, 1126–1128.

46. Guthrie,C. and Fink,G.R. (1991) Guide to Yeast Genetics and Molecular Biology. Academic Press, San Diego, CA, USA.

47. Rose,M.D., Winston,F. and Hieter,P. (1990) Methods in Yeast Genetics: A Laboratory Course Manual. Cold Spring Harbor, New York.

48. Ghazal,G., Gagnon,J., Jacques,P.E., Landry,J.R., Robert,F. and Elela,S.A. (2009) Yeast RNase III triggers polyadenylation-independent transcription termination. Mol. Cell, 36, 99–109.

49. Yen,K., Gitsham,P., Wishart,J., Oliver,S.G. and Zhang,N. (2003) An improved tetO promoter replacement system for regulating the expression of yeast genes. Yeast, 20, 1255–1262.

50. Myers,A.M., Tzagoloff,A., Kinney,D.M. and Lusty,C.J. (1986) Yeast shuttle and integrative vectors with multiple cloning sites suitable for construction of lacZ fusions. Gene, 45, 299–310.

239

51. Larose,S., Laterreur,N., Ghazal,G., Gagnon,J., Wellinger,R.J. and Elela,S.A. (2007) RNase III-dependent regulation of yeast telomerase. J. Biol. Chem., 282, 4373–4381.

52. Lamontagne,B. and Elela,S.A. (2001) Puriﬁcation and characterization of Saccharomyces cerevisiae Rnt1p nuclease. Methods Enzymol., 342, 159–167.

53. Lamontagne,B. and Abou Elela,S. (2007) Short RNA guides cleavage by eukaryotic RNase III. PLoS One, 2, e472.

54. Keogh,M.C. and Buratowski,S. (2004) Using chromatin immunoprecipitation to map cotranscriptional mRNA processing in Saccharomyces cerevisiae. Methods Mol. Biol., 257, 1–16.

55. Ashburner,M., Ball,C.A., Blake,J.A., Botstein,D., Butler,H., Cherry,J.M., Davis,A.P., Dolinski,K., Dwight,S.S., Eppig,J.T. et al. (2000) Gene ontology: tool for the uniﬁcation of biology. The Gene Ontology Consortium. Nat. Genet., 25, 25–29.

56. Lamontagne,B., Hannoush,R.N., Damha,M.J. and Abou Elela,S. (2004) Molecular requirements for duplex recognition and cleavage by eukaryotic RNase III: discovery of an RNA-dependent DNA cleavage activity of yeast Rnt1p. J. Mol. Biol., 338, 401–418.

57. Cao,H., Yue,M., Li,S., Bai,X., Zhao,X. and Du,Y. (2011) The impact of MIG1 and/or MIG2 disruption on aerobic metabolism of succinate dehydrogenase negative Saccharomyces cerevisiae. Appl. Microbiol. Biotechnol., 89, 733–738.

58. Yassour,M., Kaplan,T., Fraser,H.B., Levin,J.Z., Pﬁffner,J., Adiconis,X., Schroth,G., Luo,S., Khrebtukova,I., Gnirke,A. et al. (2009) Ab initio construction of a eukaryotic transcriptome by massively parallel mRNA sequencing. Proc. Natl Acad. Sci. USA, 106, 3264–3269.

59. Spielewoy,N., Flick,K., Kalashnikova,T.I., Walker,J.R. and Wittenberg,C. (2004) Regulation and recognition of SCFGrr1 targets in the glucose and amino acid signaling pathways. Mol. Cell. Biol., 24, 8994–9005.

60. Galdieri,L., Mehrotra,S., Yu,S. and Vancura,A. (2010) Transcriptional regulation in yeast during diauxic shift and stationary phase. OMICS, 14, 629– 638.

61. Towle,H.C. (2005) Glucose as a regulator of eukaryotic gene transcription. Trends Endocrinol. Metab., 16, 489–494.

62. Wertman,K.F., Drubin,D.G. and Botstein,D. (1992) Systematic mutational analysis of the yeast ACT1 gene. Genetics, 132, 337–350. 240

63. Larose,S. (2008), Étude des mécanismes de régulation de la télomérase chez la levure Saccharomyces cerevisiae, Ph.D. Thesis. Université de Sherbrooke, Department of Microbiology, Canada. p. 245.

64. Polish,J.A., Kim,J.H. and Johnston,M. (2005) How the Rgt1 transcription factor of Saccharomyces cerevisiae is regulated by glucose. Genetics, 169, 583–594.

65. Malagon,F., Kireeva,M.L., Shafer,B.K., Lubkowska,L., Kashlev,M. and Strathern,J.N. (2006) Mutations in the Saccharomyces cerevisiae RPB1 gene conferring hypersensitivity to 6-azauracil. Genetics, 172, 2201–2209.

66. Trumbly,R.J. (1992) Glucose repression in the yeast Saccharomyces cerevisiae. Mol. Microbiol., 6, 15–21.

67. Kuttykrishnan,S., Sabina,J., Langton,L.L., Johnston,M. and Brent,M.R. (2010) A quantitative model of glucose signaling in yeast reveals an incoherent feed forward loop leading to a speciﬁc, transient pulse of transcription. Proc. Natl Acad. Sci. USA, 107, 16743–16748.

68. Rabani,M., Levin,J.Z., Fan,L., Adiconis,X., Raychowdhury,R., Garber,M., Gnirke,A., Nusbaum,C., Hacohen,N., Friedman,N. et al. (2011) Metabolic labeling of RNA uncovers principles of RNA production and degradation dynamics in mammalian cells. Nat. Biotechnol., 29, 436–442.

69. Sabina,J. and Johnston,M. (2009) Asymmetric signal transduction through paralogs that comprise a genetic switch for sugar sensing in Saccharomyces cerevisiae. J. Biol. Chem., 284, 29635–29643.

70. McCarthy,J.E. (1998) Posttranscriptional control of gene expression in yeast. Microbiol. Mol. Biol. Rev., 62, 1492–1553.

71. Mata,J., Marguerat,S. and Bahler,J. (2005) Post-transcriptional control of gene expression: a genome-wide perspective. Trends Biochem. Sci., 30, 506–514.

72. Kushner,S.R. (2004) mRNA decay in prokaryotes and eukaryotes: different approaches to a similar problem. IUBMB Life, 56, 585–594.

73. Jacobson,A. (2004) Regulation of mRNA decay: decapping goes solo. Mol. Cell, 15, 1–2.

74. Eberle,A.B., Hessle,V., Helbig,R., Dantoft,W., Gimber,N. and Visa,N. (2010) Splice-site mutations cause Rrp6-mediated nuclear retention of the unspliced RNAs and transcriptional down-regulation of the splicing-defective genes. PLoS One, 5, e11540.

241

75. Callahan,K.P. and Butler,J.S. (2010) TRAMP complex enhances RNA degradation by the nuclear exosome component Rrp6. J. Biol. Chem., 285, 3540–3547.

76. Canavan,R. and Bond,U. (2007) Deletion of the nuclear exosome component RRP6 leads to continued accumulation of the histone mRNA HTB1 in S-phase of the cell cycle in Saccharomyces cerevisiae. Nucleic Acids Res., 35, 6268– 6279.

77. Lee,A., Henras,A.K. and Chanfreau,G. (2005) Multiple RNA surveillance pathways limit aberrant expression of iron uptake mRNAs and prevent iron toxicity in S. cerevisiae. Mol. Cell, 19, 39–51.

78. Egloff,S. and Murphy,S. (2008) Cracking the RNA polymerase II CTD code. Trends Genet., 24, 280–288.

79. Kim,M., Krogan,N.J., Vasiljeva,L., Rando,O.J., Nedea,E., Greenblatt,J.F. and Buratowski,S. (2004) The yeast Rat1 exonuclease promotes transcription termination by RNA polymerase II. Nature, 432, 517–522.

80. Johnson,S.A., Cubberley,G. and Bentley,D.L. (2009) Cotranscriptional recruitment of the mRNA export factor Yra1 by direct interaction with the 3’ end processing factor Pcf11. Mol. Cell, 33, 215–226.

81. Neil,H., Malabat,C., d’Aubenton-Carafa,Y., Xu,Z., Steinmetz,L.M. and Jacquier,A. (2009) Widespread bidirectional promoters are the major source of cryptic transcripts in yeast. Nature, 457, 1038–1042.

82. Iglesias,N., Redon,S., Pfeiffer,V., Dees,M., Lingner,J. and Luke,B. (2011) Subtelomeric repetitive elements determine TERRA regulation by Rap1/Rif and Rap1/Sir complexes in yeast. EMBO Rep., 12, 587–593.

83. Jimeno-Gonzalez,S., Haaning,L.L., Malagon,F. and Jensen,T.H. (2010) The yeast 5’-3’ exonuclease Rat1p functions during transcription elongation by RNA polymerase II. Mol. Cell, 37, 580–587.

84. Luo,W., Johnson,A.W. and Bentley,D.L. (2006) The role of Rat1 in coupling mRNA 3’-end processing to transcription termination: implications for a uniﬁed allosteric-torpedo model. Genes Dev., 20, 954–965.

85. Thomas,B.J. and Rothstein,R. (1989) Elevated recombination rates in transcriptionally active DNA. Cell, 56, 619–630.

86. Chanfreau,G., Rotondo,G., Legrain,P. and Jacquier,A. (1998) Processing of a dicistronic small nucleolar RNA precursor by the RNA endonuclease Rnt1. EMBO J., 17, 3726–3737. 242

87. Winzeler,E.A., Shoemaker,D.D., Astromoff,A., Liang,H., Anderson,K., Andre,B., Bangham,R., Benito,R., Boeke,J.D., Bussey,H. et al. (1999) Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science, 285, 901–906. 243

Supplementary Data

Regulation of conditional gene expression by coupled transcription repression and RNA degradation

Mathieu Lavoie, Dongling Ge and Sherif Abou Elela

Supplementary Table S1 : List of yeast strains used in this study

Strain Genotype Reference W303 MATa leu2-3,112 his3-11,15 trp1-1 ura3-1 ade2-1 can1-100 (85) MATa leu2-3,112 his3-11,15 trp1-1 ura3-1 ade2-1 can1-100 rnt1 rnt1::TRP1 (86) Promoter region of MIG2 replaced with ACT1 promoter in W303 PACT1-MIG2 W303 strain This study Promoter region of MIG2 replaced with ACT1 promoter in

rnt1 PACT1-MIG2 rnt1 strain This study Promoter region of RGT1 replaced with ACT1 promoter in W303 PACT1-RGT1 W303 strain This study Promoter region of RGT1 replaced with ACT1 promoter in

rnt1 PACT1-RGT1 rnt1 strain This study Promoter region of MTH1 replaced with ACT1 promoter in W303 PACT1-MTH1 W303 strain This study Promoter region of MTH1 replaced with ACT1 promoter in

rnt1 PACT1-MTH1 rnt1 stain This study W303 pMIG2pr- LacZ pMIG2pr-LacZ plasmid transformed in W303 strain This study rnt1 pMIG2pr- LacZ pMIG2pr-LacZ plasmid transformed in rnt1 strain This study W303 pRGT1pr- LacZ pRGT1pr-LacZ plasmid transformed in W303 strain This study rnt1 pRGT1pr- LacZ pRGT1pr-LacZ plasmid transformed in rnt1 strain This study W303 pMTH1pr- LacZ pMTH1pr-LacZ plasmid transformed in W303 strain This study rnt1 pMTH1pr- LacZ pMTH1pr-LacZ plasmid transformed in rnt1 strain This study xrn1 MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 xrn1Δ::KMX4 (87) rnt1/xrn1 MATa his3- leu2Δ0 ura3Δ0 lys2Δ0 rnt1Δ::HIS3 xrn1Δ::KMX4 This study rrp6Δ MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 rrp6Δ::KMX4 (48) ski7 MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ski7Δ::KMX4 (87) rat1-1 MATa his3Δ200 leu2Δ1 trp1Δ63 ura3-52 rat1-1 (48) rat1-1/xrn1 MATα ade3 his3 leu2 trp1 xrn1::URA3 rat1-1 (48)

244

Supplementary Table S2 : List of oligonucleotides used in this study

Cloning Primers Fact1-promoter GCCAGACTCGCCGGGAAGGCGGAGCTCTAACCTACATTCTTCCTTATCGGATCCTC Ract1-promoter GAGGCCGATAGGCCACTAGTGGATCTGTGTTAATTCAGTAAATTTTCGATCTTGGG Mig2-tet-for CCCCATTAAGTTAGCGTATTAACCTACAGTTAATGATTGCAGCTGAAGCTTCGTACGC Mig2-tet-rev TTTCGTTATCTACTGGGAAATTCGTTTGCTTTTTAGGCATATAGGCCACTAGTGGATCTG Rgt1-tet-for GTCACCTCACCAATAATCCGAGCTTCTTGAAATGGCAGGTCAGCTGAAGCTTCGTACGC Rgt1-tet-rev AGTCACTGGAGTTAGTCGAAACAGTGTTCAGCTCGTTCATATAGGCCACTAGTGGATCTG Mth1-tet-for TAGGATGACTGCCCCACACATTTCCTCATCTTATAATTTTCAGCTGAAGCTTCGTACGC Mth1-tet-rev CTTGGTTTTTCGAAGTTGCTGGTGGTGGTGAAACAAACATATAGGCCACTAGTGGATCTG Mig2-LacZ-F-500 CCCACCCTTTGGATCCCGATGTGCTGTAGTTGCCCT Mig2-LacZ-R GCCAAGCCGAATTCATTCTCTTTTTTTATTTTTATTTTGTTT Mth1-LacZ-F-490 CCCACCCTTTGGATCCTTATTCCGAAATTATTCCCTAGAACAA Mth1-LacZ-R GCCAAGGGGAATTCATTCCTTTGAGTGTGTGTACTCTATGC Rgt1-LacZ-F-711 CCCACCCTTTGGATCCCAACATTTGTAGTTACTTGTCTTACG Rgt1-LacZ-R GCCAAGGGGAATTCATAATTTGAAATATATTGGAGTTTGAGA ChIP Primers ChrV-For GGCTGTCAGAATATGGGGCCGTAGTA ChrV-Rev CACCCCGAAGCTGCTTTCACAATAC MTH1-5primer-1F TGCCATTCTACCCCTTATTCAAGTGCC MTH1-5primer-1R TTGCTGCTTGTGGTTTCGCTATTGAG MTH1-5primer-2F CCATCTCACGCCGAGACTTACTTGGA MTH1-5primer-2R AGATGAGGAAATGTGTGGGGCAGTCA MTH1-ORF-1F TTGTTTCACCACCACCAGCAACTTCG MTH1-ORF-1R CCTTCTGCTAGGTGAGGAGGCCAATA MTH1-ORF-2F CAGGAGATCAAGCGTGGCTGAAAGTG MTH1-ORF-2R CTTGGTAAACTTGGGACGCCCGTTAC MTH1-ORF-3F CGAGAAGTTAAGGCCCGAATGGGGTA MTH1-ORF-3R TCCGGTTTCGACAAATTCATGACGGC MTH1-LOOPF GCAGTAGAGGCACAATGCAGGTTTGA MTH1-LOOPR TGCGACCATATCGTAGCCTTTTCTTCT MTH1-3primer-1F CCAAGGAATTGAGGAAGAAGCGCCAA MTH1-3primer-1R CGGCTTCTTCTGTATGCTGTGCTCAA MTH1-3primer-2F CCACCAAAACCGATCTCGGAACATGG MTH1-3primer-2R CCCTCTTCCAAAACAAAAACACTCACTTCC PMP3-1F GCACAATACAACAATGGATTCTGCCAAG PMP3-1R TACAGTCAGTACCCCACCCACGGGC PMP3-2F GCCTTGTACATTGTCCTACAAGATTAAGCTCA PMP3-2R CCGAACATGTTGATACCGTTTCCTAATACTG Northern Blot Probes MIG2-probe-F ATGCCTAAAAAGCAAACGAATTTCC MIG2-probe-R TGGGACCGTTGAAAACATCAATTTG MTH1-Forward CAACTTCGAAAAACCAAGTTTTACAACG MTH1-Reverse GGCTGCCAATCCAATCCTAATCTTTGG RGT1-F GGGTAATATTGGCCAACAGCAGTTTTGG RGT1-R GCAACATTTACGGGGTTTGTAGATGTG LacZ2 GCAAGGCGATTAAGTTGGG MIG2-probe-F ATGCCTAAAAAGCAAACGAATTTCC Primer Extension Probes Rgt1-1545 CTGTCCATACGCGATAGTTGATGAAGTGGT MTH1-Reverse GGCTGCCAATCCAATCCTAATCTTTGG MTH5'-2end GGTGAAACAAACATTCCTTTGAGTGTGTG

245

Supplementary Figure S1. Rnt1p cleavage is required for the conditional decay of MTH1. (A) Total RNA was extracted from RNT1 and rnt1∆ cells following transcription shutdown by addition of thiolutin. Relative mRNA amount of each gene was quantitated and half-life calculated. Bar chart represents the relative increase in mRNA half-life in rnt1∆ cells compared to RNT1 cells. (B) Point mutations (boxed) where introduced in MTH1 gene to disrupt Rnt1p cleavage signal (mth1-4). Wild- type and mth1-4 cells were grown in permissive conditions (galactose) then shifted in glucose-containing media. Total RNA was extracted from cells collected at different time points and relative mRNA amount was calculated. (C) Relative Mth1 mRNA decay following a shift to restrictive conditions. RNT1 and rnt1∆ cells were grown in galactose containing media then shifted to glucose. Total RNA was extracted from cells collected at different time points and relative Mth1 mRNA amount was calculated. (D) Relative Mig2 mRNA decay following a shift to restrictive conditions. RNT1 and rnt1∆ cells were grown in glucose containing media then shifted glycerol containing media. Total RNA was extracted from cells collected at different time points and relative Mig2 mRNA amount was calculated.

246

DISCUSSION

Le contrôle précis de la synthèse des ARNs serait inutile sans un contrôle tout aussi précis de leur dégradation (Dolken et al., 2008; Elkon et al., 2010; Rabani et al., 2011). La dégradation de l’ARN se doit donc d’être hautement spécifique et régulée. Ironiquement, la majorité des ribonucléases reconnaissent des motifs génériques. Les RNase III par exemple, lient et coupent des duplexes d’ARNdb ayant peu ou pas de conservation de séquences particulières (Figure 5 de l’introduction) (Conrad et Rauhut, 2002; MacRae et Doudna, 2007). Curieusement, Rnt1p retrouvé chez Saccharomyces cerevisiae semble faire exception à cette règle. En effet, la comparaison des substrats connus jusqu’ici ainsi que de nombreux essais biochimiques et structuraux ont démontrés que Rnt1p reconnaît spécifiquement des tétra-boucles « NGNN » (Ghazal et Abou Elela, 2006; Ghazal et al., 2005; Lamontagne et al., 2004). De plus, il a été observé que la délétion de RNT1 cause des effets pléiotropiques sur l’expression de nombreux gènes codants et non-codants (Ge et al., 2005; Lee et al., 2005), suggérant que Rnt1p est un important régulateur post-transcriptionnel de l’expression génique.

Au cours de cette thèse, je me suis donc intéressé aux mécanismes de reconnaissance de l’ARN et aux rôles fonctionnels de Rnt1p afin de mieux comprendre comment l’expression des gènes est régulée chez la levure. Notamment, les résultats ont permis de mettre en lumière les règles qui dictent la spécificité de Rnt1p pour les tétra-boucles NGNN. En outre, un nouveau motif structurel, unique à Rnt1p, situé à l’extrémité du dsRBD assure la sélection positive de la tétra-boucle tout en limitant la reconnaissance des hélices d’ARNdb génériques. En parallèle, j’ai developpé des nouvelles approches biochimiques qui ont permis d’identifier plusieurs centaines de nouveaux transcrits directement coupés par Rnt1p et dont la majorité sont des ARNm. La plupart de ces ARNm sont régulés par Rnt1p de manière dépendante des conditions de culture de la 247

levure, démontrant ainsi le rôle important de la dégradation nucléaire dans la régulation conditionnelle de l’expression des gènes chez les eucaryotes.

6. La reconnaissance des substrats par Rnt1p: entre flexibilité et spécificité

6.1. Rnt1p reconnaît la structure et la séquence de la tétra-boucle

En 2004, la détermination de la structure en solution du dsRBD de Rnt1p en complexe avec une tétra-boucle AGAA a permis d’établir un modèle selon lequel la reconnaissance des substrats par Rnt1p repose strictement sur la structure de l’ARN et non sur sa séquence (Figure 6) (Wu et al., 2004). Ce modèle est encore présenté à ce jour dans les articles scientifiques pour expliquer la spécificité de Rnt1p pour les tétra-boucles NGNN (Thapar et al., 2014). Toutefois, ce modèle partiel n’explique pas comment le site de clivage est choisi par rapport à la tétraboucle, ni comment Rnt1p reconnaît des tétra-boucles non canoniques AAGU et la contribution du NTD dans ce processus (Gaudin et al., 2006; Ghazal et Abou Elela, 2006). Dans le but de raffiner et de compléter ce modèle, j’ai donc voulu caractériser les déterminants chimiques et structuraux de l’ARN et de la protéine essentiels pour la liaison et le clivage par Rnt1p.

Puisque la structure de l’ARN ne semble pas être le seul déterminant pour la reconnaissance par Rnt1p, il a été proposé que Rnt1p forme des interactions spécifiques avec l’ARN pour lier et cliver ses substrats. Afin d’identifier les interactions critiques pour la liaison et la coupure, j’ai d’abord utilisé des nucléotides modifiés dont le groupement 2’-hydroxyl du ribose a été substitué par des groupements 2’-O-méthyl (chapitre I) (Lavoie et Abou Elela, 2008). Dans l’ensemble, toutes les modifications 2’-O-methyl au niveau de la tétra-boucle ont abolit le clivage, ce qui indique que Rnt1p demeure très sensible aux changements de structure de l’ARN. Afiin de contourner ce problème, nous avons utilisé des modifications 2’-fluoro qui n’ont pas la capacité de former de ponts hydrogènes avec la protéine tout en préservant la structure de l’ARN (Egli et Gryaznov, 2000). Les modifications effectuées obtenus montrent qu’aucune des interactions n’est essentielle à la liaison et le clivage par Rnt1p. Toutefois, certaines modifications, 248

notammant au niveau de la guanosine conservée, ont démontré un impact significatif sur la liaison de l’enzyme. En conclusion, Rnt1p reconnait la structure de la tétra-boucle NGNN par la formation d’un réseau redondant de ponts hydrogènes.

La détermination de la structure cristaline de Rnt1p en complexe avec son substrat a permis de confirmer la majorité des résultats obtenus avec les modifications 2’fluoro. En effet, la structure cristalline a révélé la formation de ponts hydrogène pour la majorité des groupements 2’-hydroxyl de l’ARN pour lesquels la substitution pour des groupements 2’-fluoro causait une diminution de la liaison ou du clivage in vitro (chapitres I et II). Ces deux projets ont également permis de confirmer les résultats de Wu et collègues, c’est-à-dire que la liaison du dsRBD à l’ARN requiert la participation de nombreux ponts hydrogènes avec les nucléotides en position 3 et 4 de la tétra-boucle et des paires de bases adjacentes (Wu et al., 2004). D’un autre côté, la réalisation des essais avec la protéine complète a permis d’identifier de nombreuses interactions additionnelles entre la Rnt1p et la tétra-boucle, notamment avec le nucléotide à la première position et la guanosine conservée. Plus précisément, mes travaux ont révélé l’existance d’un nouveau motif situé à l’extrémité C-terminale du dsBRD, que nous avons appelée « G clamp », qui est essentiel pour la liaison et le clivage de l’ARN (Liang et al., 2014). Ce motif sert à la reconnaissance spécifique de la guanosine conservée en formant plusieurs interactions avec le sucre, le phosphate et la base elle-même (chapitre II). Ainsi, l’identité de la base à la deuxième position de la tétra-boucle semble jouer un rôle important pour assurer la spécificité de Rnt1p en formant des interactions spécifiques avec Rnt1p et en assurant le repliement adéquat du squelette de la tétra-boucle.

La structure cristalline de Rnt1p en complexe avec son produit de clivage a aussi montré que le NTD contacte les nucléotides du côté 5’ de la tétra-boucle (chapitre II). Le NTD formerait au moins trois ponts hydrogènes avec les groupements phosphate du squelette de l’ARN et un pont hydrogène avec la guanosine conservée. Ce résultat est en accord avec ceux obtenus à partir des 249

substrats comprenant des ribonucléotides modifiés (chapitre I), ainsi qu’avec les essais d’empreintes aux radicaux hydroxyl et de résonance magnétique nucléaire qui suggèrent tous que Rnt1p interagit avec le côté 5’ de la tétra-boucle d’une manière dépendante de la présence du NTD (Ghazal et Abou Elela, 2006; Lamontagne et al., 2003; Lavoie et Abou Elela, 2008). Il corrèle également avec le fait que le NTD soit requis pour la stabilisation du complexe enzyme:substrat (Lamontagne et al., 2000). Bien que la structure cristalline n’ait pas détecté la formation d’un pont hydrogène entre le NTD et les groupements 2’-hydroxyl de l’ARN, tel que suggéré au chapitre I, il est possible que le NTD participe à la formation ou à la stabilisation des ponts hydrogènes avec le dsRBD à cet endroit. Ainsi, le modèle général proposé est que le dsRBD est suffisant pour lier le coté 3’ de la tétra-boucle, mais que les interactions sont stabilisées par la présence du NTD du côté 5’ (Ghazal et Abou Elela, 2006). Ainsi, ces traveaux ont permis de mieux comprendre le rôle du domaine NTD dans la sélection et le clivage des substrats par Rnt1p.

Une conclusion importante qui découle de l’ensemble de ces résultats est que la reconnaissance des tétra-boucles NGNN par Rnt1p ne repose pas seulement sur la structure particulière adoptée par l’ARN, mais aussi sur sa séquence. En effet, la base conservée est contactée par deux domaines de Rnt1p, soit le motif G-clamp situé dans l’extension C-terminale de la protéine et le NTD.

6.2. La reconnaissance des substrats non-canoniques nécessite vraisemblablement un changement conformationnel de l’ARN

La présence de la « G clamp » soulève une importante question : comment Rnt1p reconnaît-il d’autres structures d’ARN? En effet, Rnt1p lie et clive des tétra- boucles AAGU qui n’ont pas de guanosine à la seconde position et qui, en solution, adoptent une structure différente de la tétra-boucle NGNN (Gaudin et al., 2006; Ghazal et Abou Elela, 2006). Une étude précédente a démontré que la liaison du dsRBD entraine un changement de conformation de l’ARN qui fait en sorte que sa structure imite celle de la tétra-boucle NGNN (Wang et al., 2011). 250

Parallèlement, nous avons observé que les mutations effectuées dans la « G clamp » affecte le clivage des substrats AAGU d’une manière similaire aux substrats NGNN, suggérant que ce motif participe aussi à la liaison des tétra- boucles AAGU (Mathieu Lavoie, données non publiées). En théorie, la « G clamp » est capable d’accommoder l’adénosine à la seconde position, mais l’interaction serait moins stable en raison de l’absence de certains ponts hydrogènes. Par ailleurs, il faut se rappeller que les substitutions 2’-O-methyl ont démontré que Rnt1p demeure très sensible à la modification de la structure. Ainsi, considérant l’ensemble des résultats obtenus, je crois que la reconnaissance des tétra-boucles par Rnt1p repose principalement sur un changement conformationnel de l’ARN qui imite la structure de la tétra-boucle NGNN et permet de positionner l’adénosine à la seconde position au niveau de la G-clamp. Ce modèle n’exclut pas un possible ré-arrangement du dsRBD de Rnt1p étant donné que la flexibilté relative de mouvement de l’hélice α3 qui influence le positionnement de l’hélice α1 (Wang et al., 2011). Ces changements conformationnels de la protéine et de l’ARN expliqueraient le changement de ponts hydrogènes requis pour la liaison des tétra- boucles AAGU (chapitre I). Une autre possibilité, moins probable, est que le que la guanosine en troisième position pourrait être reconnue par la « G clamp » suite à un changement de conformation de l’ARN et/ou de la protéine. Dans tous les cas, la résolution de ce dilemme nécessitera de déterminer la structure de la protéine Rnt1p entière en complexe avec une tétra-boucle AAGU.

Bien que Rnt1p démontre une spécificité élargie pour les tiges-boucles ayant des structures différentes, il n’est pas en mesure de cliver des tétra-boucles de type GNRA (Chanfreau et al., 2000). Or, la base à la deuxième position de la tétra-boucle GAAA ne s’empile pas sur celle à la première position comme dans le cas des NGNN, mais plutôt avec les bases aux positions 3 et 4 (Correll et Swinger, 2003; Thapar et al., 2014). De plus, le G en première position et le A en dernière position forment une paire de base qui modifie la structure du squelette de l’ARN et prévient la formation du sillon mineur reconnu par l’hélice α1 de Rnt1p (Correll et Swinger, 2003; Thapar et al., 2014). En conclusion, les tétra-boucles GNRA 251

forment une structure rigide et trop différente pour qu’elle soit reconnue efficacement par Rnt1p.

Nous avons aussi observé que Rnt1p est capable de couper des tri-boucles et des penta-boucles (chapitre III). Le mécanisme permettant la reconnaissance de ces substrats n’a pas été élucidé. Il sera particulièrement pertinent d’effectuer des essais de protections aux modifications chimiques afin de déterminer la stabilité de la boucle ainsi que les positions contactées par Rnt1p. Naturellement, l’hypothèse envisagée est que ces substrats adoptent également une structure similaire à la tétra-boucle NGNN suite à un changement conformationnel de l’ARN. Ils témoignent néanmoins de la capacité de Rnt1p à reconnaître un éventail relativement large de structures tige-boucle, excluant les tétra-boucles GNRA.

Dans l’ensemble, les résultats obtenus avec les substitutions de groupements 2’-OH et les mutations ponctuelles de Rnt1p démontrent que la majorité des interactions entre la protéine et l’ARN sont redondantes, c’est-à-dire que l’abolition individuelle des liens n’a généralement qu’un effet limité sur le clivage. C’est possiblement cette redondance qui procure la flexibilité à l’enzyme pour reconnaître différentes structures d’ARN. En conséquence, il est possible d’envisager un modèle selon lequel Rnt1p distingue ses substrats en fonction du nombre d’interactions créées avec l’ARN et de la rapidité à former celles-ci. Ainsi, Rnt1p aurait la possibilité de sonder rapidement de nombreuses structures d’ARN et uniquement celles qui forment suffisamment de contacts, comme dans le cas des tétra-boucles NGNN, seront liées assez longtemps pour être coupées.

6.3. Rnt1p révèle un nouveau mode de reconnaissance de l’ARN par les RNase III

Le mécanisme de reconnaissance des substrats par la RNase III bactérienne a été largement étudié (Gan et al., 2008). À l’opposé, celui des RNase III eucaryotes est encore mal compris. La détermination de la structure cristalline de Rnt1p en complexe avec son substrat a permis pour la première fois de disséquer en détail le mode de reconnaissance des RNase III eucaryotes. En 252

excluant le NTD, Rnt1p lie ses substrats en utilisant cinq motifs de liaison à l’ARN, ou RBM (« RNA binding motifs ») (chapitre II). Sur l’ARN, ces motifs contactent les trois boîtes (IBPB, BSB et CEB) connues pour influencer la liaison et le clivage (voir figure 5A), en plus d’une nouvelle boîte MB située à environ neuf paires de bases sous de la tétra-boucle qui est contactée par le dsRBD et le RIIID (Figure 7) (Lamontagne et al., 2003; Liang et al., 2014). Tel que présenté à la figure 7 du chapitre II, on note généralement une très bonne corrélation entre le positionnement des RBMs, les ponts hydrogènes prédits par les 2’-Fluoro (chapitre I) ou observés dans la structure en solution du dsRBD (Wu et al., 2004)) ainsi que les nucléotides protégés de la modification chimique lors de la liaison de Rnt1p (Ghazal et Abou Elela, 2006). Par exemple, des interactions entre la boîte MB de l’ARN et la protéine ont aussi été observés lors des essais d’empreinte aux radicaux hydroxyl (Ghazal et Abou Elela, 2006). Par ailleurs, les mutations ponctuelles des acides aminés situés dans les RBM0 et 2 montrent que ces régions sont importantes pour la formation et la stabilité du complexe enzyme : substrat. En conclusion, Rnt1p interagit avec son substrat via cinq motifs de liaison qui sont distribués sur l’ensemble de la tige-boucle, du sommet de la boucle jusqu’en aval des sites de coupures.

Figure 7 : Rnt1p contacte la tétra-boucle et l’hélice d’ARNdb. Les RBMs qui contactent l’ARN sont représentés par des ovales et colorés selon les domaines et les sous-unités de Rnt1p auxquels ils appartiennent. Les nucléotides qui montrent un enrichissement statistiquement significatif (valeur p <0.05) à travers l’ensemble des substrats identifiés sont indiquées par leur lettre correspondante. « N » signifie aucun enrichissement. Similairement, les points noirs montrent les positions qui sont significativement plus appariées. Les quatre régions de l’ARN qui sont contactés par la protéine (IBPB, BSB, MD et CEB) sont encadrées. Voir le texte principal pour plus de détails.

253

En dépit de leur positionnement relatif différent, quatre de ces RBMs sont analogues à ceux retrouvés chez la RNase III bactérienne ce qui suggère que le mécanisme basal de reconnaissance de la structure de l’ARN est conservé entre les enzymes de la classe I et II (chapitre II) (Gan et al., 2006). Tel que vu en introduction, la RNase III bactérienne est capable de cliver efficacement des simples hélices d’ARNdb de forme A (Lamontagne et Abou Elela, 2004). La liaison stable requiert la présence d’ions magnésium (Li et Nicholson, 1996). De plus, le RIIID de la RNase III bactérienne est capable à lui seul de cliver de manière spécifique un ARNdb même en absence du dsRBD. Ainsi, le mécanisme proposé chez la bactérie est que le dsRBD lie la majorité des longs ARNdb alors que la spécificité est assurée par le RIIID et par la présence d’antidéterminants sur le substrat. À l’opposé, Rnt1p peut lier efficacement son substrat en absence d’ions magnésium mais la liaison stable requiert à la fois la présence du dsRBD et d’une tétra-boucle NGNN (Lamontagne et Abou Elela, 2004). Puisque que le RIIID de Rnt1p seul n’est pas capable de lier et cliver efficacement les substrats et que le clivage en absence d’une tétra-boucle est généralement non-spécifique, il semble donc que c’est le dsRBD qui assure la la spécificité de la reconnaissance (Lamontagne et Abou Elela, 2004). En effet, nous avons démontré que Rnt1p possède un motif additionnel RBM0 qui assure la reconnaissance d’un motif structurel spécifique. Bref, malgré la conservervation des RBMs 1 à 4, les enzymes procaryotes et eucaryotes ont évolué des mécanismes distincts de sélection des substrats. En effet, la RNase III bactérienne utilise la reconnaissance générale de l’hélice de forme A de l’ARNdb alors que Rnt1p utilise mécanisme de sélection positive d’un motif particulier tout en évitant la reconnaissance de l’hélice d’ARNdb générique de forme A.

Une des hypothèses soulevées pour expliquer l’acquisition d’un tel mode de reconnaissance par Rnt1p est que la levure devait empêcher la coupure de l’ARNdb afin de maintenir la survie du virus killer (Drinnenberg et al., 2011). En effet, ce virus qui procure un avantage évolutif à la levure S. cerevisiae, possède un génome d’ARNdb. Or, tant chez l’humain (Dicer) que chez la bactérie, les 254

RNase III sont impliquées dans la défense anti-virale et la destruction des génomes d’ARNdb exogènes. Dans le cas de S. cerevisiae, on peut penser que la sélectivité particulière de Rnt1p a évoluée non pas pour reconnaître un épitope précis, mais plutôt pour prévenir la coupure d’ARNdb associée à la défense anti- virale.

6.4. Un mécanisme d’action séquentiel assure la spécificité du clivage par Rnt1p

L’ensemble des données de la littérature et les résultats obtenus au cours de ce projet de recherche m’ont permis de proposer un mécanisme d’action pour la reconnaissance des tétra-boucles NGNN par Rnt1p (Figure 8). La réaction de clivage débute par l’arrimage d’un homo-dimère de Rnt1p à la tétra-boucle NGNN. Cette interaction nécessite l’interaction de la « G clamp » avec le second nucléotide de la boucle. Simultanément, ou peu après, les autres nucléotides de la tétra-boucle (IBPB) ainsi que les paires de bases adjacentes (BSB) sont liées par les RBMs 0 et 1 du dsRBD. Cette étape permet d’assurer la liaison stable du substrat. Cette étape est supportée par le fait que Rnt1p, tout comme la RNase III bactérienne, est capable de lier un ARN de manière stable sans clivage (Blaszczyk et al., 2004; Lamontagne et Abou Elela, 2004). À ce stade, un appariement inadéquat pourrait mener à la formation d’un complexe stable, mais non productif. En effet, certaines substitutions des groupements 2’-OH au niveau de la tétra- boucle semblaient causer la formation d’un tel complexe non-productif (Lavoie et Abou Elela, 2008). L’étape suivante de chargement permet de positionner le substrat au niveau de la vallée catalytique du RIIID et repose essentiellement sur la liaison de la boîte MB par les RBMs 2 et 4 et possiblement de la boîte CEB par le RBM 3 du RIIID. Un mécanisme similaire a été observé dans le cas de la RNase III bactérienne (Gan et al., 2005). L’étape suivante de verrouillage est représentée par la liaison du dimère de NTD du côté 5‘ de la boucle. Même si non essentielle, cette étape permet d’assurer la stabilité du complexe enzyme:substrat en renforçant notamment les interactions entre l’ARN et le dsRBD. L’absence du NTD se caractérise par une augmentation de l’impact causé par les substitutions 2’-OH 255

(chapitre I) et par une diminution de la stabilité du complexe en fonction de la concentration en sels monovalents (Lamontagne et al., 2000). Par ailleurs, des essais de double-hybride ont montré que le NTD et le dsRBD étaient capables d’interagir entre eux (Lamontagne et al., 2000). La liaison du NTD favorise aussi le positionnement adéquat des centres catalytiques du dimère de RIIID avec les sites de coupures respectifs. Ainsi, le NTD constituerait un second système de mesure qui permet d’assurer la précision du site de coupure (chapitre II). Suivra ensuite l’étape irréversible de la coupure des liens phosphodiesters de l’ARN qui nécessite l’apport d’ions magnésium et la coordination de six acides aminés conservés au niveau du cœur catalytique (Liang et al., 2014). Finalement, il y aura le relâchement des produits et la dissociation du complexe. Cette dernière étape peut être plus ou moins rapide en fonction de la présence de paires de bases au niveau du CEB (Comeau et Abou Elela, manuscrit en préparation). Bien que supporté par de nombreuses données biochimiques et génétiques, ce modèle demeure hypothétique puisque seulement la structure du complexe post-catalytique a été observée jusqu’ici. Des essais supplémentaires réalisés avec différents mutants de Rnt1p devront être effectués pour confirmer chacun des états intermédiaires du mécanisme. En conclusion, le mécanisme d’action de Rnt1p nécessite plusieurs étapes séquencielles qui impliquent la formation de complexes enzyme : subtrats catalytiques et non-catalytiques et qui, ensemble, assurent la spécificité du site de clivage. 256

Figure 8 : Mécanisme d’action de Rnt1p. Rnt1p fonctionne en tant que dimère antiparallèle. Les trois principaux domaines de Rnt1p sont représentés par des cylindres reliés par des jonctions flexibles et colorés en fonction de la sous-unité respective (voir la légende). Les numéros représentent les RBMs potentiellement impliqués à chaque étape. Voir le texte principal pour plus de détails.

6.5. Différents éléments cis influencent la réactivité des substrats de Rnt1p

Jusqu’à maintenant, la définition d’un substrat de Rnt1p se limitait à une tétra-boucle NGNN suivi de cinq paires de bases (Lamontagne et Abou Elela, 2004; Lamontagne et al., 2003). L’identification de nouvelles cibles de Rnt1p au cours de cette étude a permis de mieux définir la structure et la séquence consensus d’un substrat type (Figure 7). Tel qu’attendu, les sept premières paires de bases situées sous la tétra-boucle sont généralement appariées, assurant ainsi la formation d’une tétra-boucle stable. La séquence consensus de la tétra-boucle semble être AGDU, où le D indique une sous-représentation statistiquement significative de la cytosine. Soulignons que cette séquence consensus n’est pas absolue puisque Rnt1p clive très bien des tétra-boucles dont la séquence diffère de celle-ci (par exemple, snR64 et snR56 ont des tétra-boucles AGCA et UGGU, respectivement) (Ghazal et al., 2005; Lee et al., 2003). On peut donc penser que les éléments présentés dans le nouveau modèle consensus sont ceux qui sont les plus susceptibles d’influencer positivement la réactivité des substrats. La raison de 257

l’exclusion de la cytosine à la position 3 n’est pas connue, mais on peut suggérer que sa présence ne favorise pas le repliement optimal de la structure, empêchant ainsi la formation du lien entre Rnt1p et le groupement phosphate. En effet, tel que mentionné en introduction, les bases aux positions 3 et 4 de la tétra-boucle forment un empilement avec la base adjacente en 3’, ce qui contribue à stabiliser la structure (Lebars et al., 2001). La préférence pour une adénosine à la première position a déjà été observée (Lamontagne et al., 2003). Quelques positions dans la tige montrent aussi un enrichissement significatif de séquence, notamment la cytosine immédiatement en 5’ de la tétra-boucle. Ce résultat contredit l’observation précédente selon laquelle la présence d’une paire de base C-G, plutôt que G-C, à cet endroit réduisait fortement le clivage et la liaison de l’enzyme, spécialement en présence de courtes hélices d’ARNdb (Lamontagne et al., 2003). Toutefois, tel que mentionné par les auteurs, cette inhibition semblait dépendante de l’identité des séquences adjacentes. Ainsi, la reconnaissance des substrats par Rnt1p semble reposer sur des combinaisons spécifiques de séquence et de structure de l’ARN.

Deux exemples illustrent bien l’impact des séquences présentes dans la tige d’ARNdb sur la réactivité des substrats. Premièrement, il a déjà été démontré que la présence de paires de bases G-C au site de clivage diminue le clivage, suggérant un mécanisme analogue aux anti-déterminants qui régulent la RNase III bactérienne (Lamontagne et al., 2003; Zhang et Nicholson, 1997). Justement, nous avons pu observer dans la structure cristaline de Rnt1p que la région CEB est contactée par les RBM 3 respectifs du dimère de RIIID. De plus, des résultats obtenus au laboratoire montrent que la présence des paires de bases immédiatement en amont des sites de clivage diminue le taux de rotation (Kcat) de Rnt1p, vraisemblablement par un mécanisme d’inhibition par le produit (Comeau et Abou Elela, manuscrit en préparation). Un mécanisme similaire d’inhibition par le produit a déjà été proposé dans le cas de la RNase III bactérienne (Campbell et al., 2002). Finalement, il est aussi démontré que le relâchement des produits constitue une étape limitante lors du clivage par l’enzyme Dicer de Kluyveromyces 258

polysporus (Weinberg et al., 2011). L’inhibition par le produit pourrait constituer un mécanisme ingénieux de rétrorégulation de l’activité de Rnt1p in vivo.

Un deuxième exemple d’épitope régulant la réactivité des substrats est la boîte MB qui est contactée par le dsRBD et le RIIID. Les résultats obtenus démontrent que l’interaction entre la MB et le RBM 2 semble requise pour la liaison optimale et la catalyse (chapitre II). Or, l’examen des nouveaux substrats identifiés au chapitre III (Gagnon et al., 2014) révèle que cette région de l’ARN est généralement moins appariée que le reste de la tige (Figure 7). Ainsi, la présence d’une boucle interne (« bulge ») à cet endroit pourrait être un moyen de réguler l’activité de l’enzyme.

De toute évidence, Rnt1p ne coupe pas systématiquement toutes les tétra- boucles avec la même efficacité in vitro. Cette préférence intrinsèque pour certaines tétra-boucles in vitro se reflète généralement aussi in vivo. En effet, les tige-boucles isolées à partir de transcrits qui ne sont pas coupés par Rnt1p (tel que l’ARNm du gène POM33) montrent souvent un taux de clivage faible et peu efficace in vitro (chapitre III). De plus, l’insertion de sites de coupure dérivés de substrats connus de Rnt1p dans un gène rapporteur a démontré que l’accumulation de produits de clivage in vivo est corrélée avec la capacité de Rnt1p à couper ces tiges-boucles in vitro (chapitre IV) (Meaux et al., 2011). Similairement, le groupe de Christina Smolke a inséré dans un gène rapporteur une librairie de tige-boucles synthétiques clivées plus ou moins efficacement par Rnt1p (Babiskin et Smolke, 2011b, c). Ils ont ainsi observé une corrélation entre le taux de clivage par Rnt1p in vitro et la répression du transcrit in vivo. Ils concluent également que la préférence intrinsèque de Rnt1p pour certaines tétra-boucles in vitro et in vivo provient d’une combinaison de séquence et de structure de l’ARN au niveau des boîtes BSB et CEB connues pour influencer la liaison et le clivage, plutôt que d’une caractéristique unique et/ou spécifique à chaque transcrit.

Outre la tétra-boucle elle-même, d’autres éléments cis tels que les séquences d’ARN adjacentes et les régions promotrices semblent aussi réguler 259

l’activité de Rnt1p. En effet, les séquences adjacentes peuvent entre autre favoriser la formation de la structure tige-boucle, comme dans le cas du transcrit polycistronique snR67/snR53 dont le site de coupure est formé grâce à des interactions à longue distance dans l’ARN (Ghazal et al., 2005). Elles peuvent également servir de site de liaison pour d’autres protéines qui facilitent le recrutement de Rnt1p (voir à la section 8.1. ci-bas) ou alors bloquer l’accès de Rnt1p à la tétra-boucle lors du repliement tertiaire de l’ARN. Parallèlement, le remplacement de la région promotrice de deux substrats de Rnt1p, MIG2 et MTH1, par le promoteur du gène ACT1 a diminué l’impact sur l’expression observé lors de la délétion de RNT1 (chapitre 5). La région promotrice est susceptible d’influencer le recrutement de Rnt1p à ses substrats (voir à la section 8.2.) ou alors le repliement de l’ARN en cours de transcription. En conclusion, le contexte génétique dans lesquels sont placées les tétra-boucles influence leur réactivité.

Ce phénomène particulièrement évident dans le cas de snR47. Normalement, le précurseur de snR47 est clivé très efficacement in vivo (Chanfreau et al., 1998a) et in vitro (chapitre III). Par contre, l’insertion de la tétra- boucle snR47 dans un contexte hétérologue semble avoir complètement abolit le clivage in vivo (chapitre IV). Une possibilité évoquée pour expliquer ce résultat est que la structure tige-boucle n’est pas bien formée dans le contexte du gène rapporteur et en présence de conditions physiologiques. Alternativement, il est possible que le clivage dépende de facteurs uniquement retrouvés dans le contexte biologique original. En appui à cette hypothèse, des résultats préliminaires d’immunoprécipitation de la chromatine indiquent que la délétion de RNT1 affecte la transcription par l’ARN polymérase II au locus de snR47 (voir l’annexe 1 et la section 7.4. ci-bas) et que Rnt1p s’associe fortement avec l’ADN du gène d’une manière dépendante de l’ARN (voir l’annexe 2). Ces résultats suggèrent que le clivage par Rnt1p est co-transcriptionnel, donc dépendant de la composition du complexe transcriptionnel et/ou de la région promotrice spécifique à snR47. Dans tous les cas, cet exemple illustre bien l’impact du contexte biologique sur la réactivité des tétra-boucles de Rnt1p. 260

La majorité des ARNs pour lesquelles une structure tétra-boucle a été prédite ne sont pas coupés par Rnt1p in vitro (chapitre III). Il est proposé qu’une combinaison inadéquate des éléments au niveau de la tétra-boucle et/ou un contexte biologique défavorable explique l’absence de réactivité des substrats. Néanmoins, en considérant l’omniprésence des tétra-boucles NGNN dans le génome de la levure (chapitre III) ainsi que la flexibilité de Rnt1p pour la reconnaissance de ses substrats (chapitre I), on réalise que l’activation ou l’inhibition du clivage de plusieurs transcrits repose potentiellement sur bien peu de modifications ou de mutations au sein du transcrit.

7. L’impact fonctionnel de Rnt1p sur l’expression des gènes

7.1. Rnt1p régule l’expression et clive plusieurs ARNm

Chez E. coli, on estime qu’environ 9% des gènes sont surexprimés suite à l’inactivation de la RNase III (Stead et al., 2011). Similairement, nous avons observé que plus de 721 transcrits, soit environ 10% des gènes chez S. cerevisiae, sont surexprimés plus de deux fois en absence de RNT1 (chapitre III). Bien que la mesure des niveaux d’expression des ARNm soit couramment utilisée afin d’identifier les cibles de ribonucléases, un problème majeur avec cette approche est qu’elle ne permet pas de d’établir un lien direct entre la surexpression et la dégradation spécifique d’un transcrit par une ribonucléase. Afin de résoudre ce problème, j’ai développé deux nouvelles techniques permettant l’identification des cibles directes de Rnt1p in vitro à l’échelle du génome. Ces techniques ont permis d’identifier plus de 384 transcrits coupés par Rnt1p in vitro (résumés en Annexe 3). De ce nombre, 329 sont des nouvelles cibles dont la majorité représente des ARNm ayant un site de coupure situé à l’intérieur de la séquence codante. Ce résultat suggère que le rôle de Rnt1p dans le clivage des ARNm est beaucoup plus important qu’initialement établit. 261

Une question évidente est de savoir si les cibles identifiées in vitro sont véritablement clivées par Rnt1p in vivo. Une combinaison de techniques et de conditions de culture a permis de démontrer que la délétion de RNT1 influence le niveau d’expression in vivo de la majorité des cibles (chapitre III). Bien que ces résultats ne constituent pas une preuve irréfutable, ils suggèrent que Rnt1p régule l’expression de la plupart de ces gènes en coupant leur ARNm in vivo. Une des méthodes la plus efficace pour démontrer l’impact direct du clivage sur la régulation in vivo est d’insérer des mutations silencieuses dans l’ARN afin de détruire le site de reconnaissance de Rnt1p sans affecter l’expression de l’ARNm ou de la protéine. La régulation directe de Rnt1p sur l’ARNm de EST1 (Larose et al., 2007), MIG2 (Ge et al., 2005), BDF2 (Roy et Chanfreau, 2014), du transcript polycistronique NPL3-GPI17 (Ghazal et al., 2009) et MTH1 (chapitre V) (Lavoie et al., 2011) a pu être démontrée grâce à cette technique. Malheureusement, il est difficile et laborieux de procéder à la mutation systématique de chaque tige-boucle. D’un autre côté, plusieurs des nouveaux gènes identifiés in vitro sont impliqués dans des voies métaboliques pour lesquelles l’implication de Rnt1p a été démontrée in vivo, telles que la division cellulaire, la réduction de la production des ribosomes, le stress des parois cellulaires ou la réponse au glucose (Abou Elela et al., 1996; Catala et al., 2012; Catala et al., 2004; Ge et al., 2005). Considérant l’ensemble de ces résultats, il est raisonnable de croire que la majorité des cibles identifiées in vitro sont effectivement régulées par Rnt1p dans la cellule. Naturellement, il n’est pas exclu que certains substrats ne soient jamais clivées directement par Rnt1p in vivo, soit parce que la structure tige-boucle n’est pas formée ou bien parce que celle-ci est masquée par d’autres protéines ou alors parce que l’ARNm est rapidement exporté au cytoplasme.

Le clivage d’ARNm par des RNase III n’est pas unique à Rnt1p. Chez les bactéries, on recense à ce jour environ une dizaine d’ARNm dont le clivage direct par la RNase III a été démontré (Court et al., 2013; Lim et al., 2012). De son côté, Drosha lie et clive des structures tige-boucle retrouvés dans au moins six ARNm, dont celui qui code pour sa protéine associée DGCR8 (Chong et al., 2010; 262

Johanson et al., 2013). Plusieurs structures tiges-boucles ressemblant à des pré- microARN ont été identifiées dans des ARNm chez l’humain. De plus, environ 12% des microARNs proviendrait de séquences exoniques de gènes codants et non- codant (Chong et al., 2010). En ce qui concerne Rnt1p, l’ensemble des candidats identifiés au cours de cette étude constitue sans doute le plus large recensement d’ARNm clivés par une RNase III in vitro et illustre l’immense potentiel des RNase III pour réguler l’expression des ARNm. Il serait intéressant d’adapter les méthodes développées au cours de ce projet à d’autres organismes et classes de RNase III.

7.2. La dégradation nucléaire est impliquée dans régulation conditionnelle de l’expression génique

Tel que démontré au chapitre IV, le clivage des ARNm par Rnt1p génère des produits instables qui sont exportés au cytoplasme où ils sont rapidement dégradés par Xrn1p et l’exosome (Meaux et al., 2011). De plus, l’absence de coiffe ou de queue poly-A réduit grandement la traduction des produits de clivage (Meaux et al., 2011; Meaux et Van Hoof, 2006). Ainsi, la dégradation nucléaire par Rnt1p constitue un système efficace de régulation de l’expression génique puisque celui-ci coupe les transcrits indésirables avant leur export au cytoplasme, prévenant ainsi leur traduction accidentelle.

De toute évidence, il serait contre-productif pour la cellule de produire un ARNm qui serait systématiquement et inconditionnellement coupé par Rnt1p. En outre, plusieurs expériences suggèrent que le clivage de certains ARNm par Rnt1p est conditionnel. Par exemple, il a été observé au chapitre III que l’accumulation de plusieurs ARNm identifiés comme substrats de Rnt1p était dépendante de la modification des conditions de culture. Il a aussi été démontré que Rnt1p accélère la répression de l’expression de MIG2 (Ge et al., 2005) et de MTH1 (chapitre V) lors du retrait ou de l’ajout de glucose dans le milieu, respectivement. Similairement, la délétion de RNT1 retarde la répression de l’expression de HSL1 en fonction du cycle cellulaire (Catala et al., 2012). En conclusion, Rnt1p participe à la répression conditionnelle de plusieurs gènes en clivant leur ARNm in vivo. 263

Tant chez l’humain que chez la levure, la modulation de la stabilité de l’ARN joue un rôle important dans la régulation de l’expression génique en réponse à un stress (Rabani et al., 2011; Shalem et al., 2008). La dégradation de l’ARN participe notamment à la coordination de l’expression génique et affecte la cinétique de répression et d’induction des gènes (Elkon et al., 2010; Romero-Santacreu et al., 2009). En ce qui concerne Rnt1p, au moins trois modèles de régulation de l’expression génique ont été observés (Figure 9). Dans le premier modèle, Rnt1p réprime conditionnellement l’expression des gènes de manière concomitante, ou même avant, l’inhibition de la transcription (Figure 9A). Ainsi, la dégradation de l’ARN augmente la rapidité et l’acuité de la répression génique. Ce modèle suggère que Rnt1p est activement et conditionnellement recruté sur ses cibles. Les ARNm MIG2 et MTH1 semblent être régulés selon ce modèle. Alternativement, il est possible d’imaginer un gène pour lequel le taux de transcription demeure constant et c’est le recrutement ou l’inhibition conditionnelle de Rnt1p qui assure la modulation des niveaux d’expression (Figure 9B). Ce modèle serait particulièrement utile dans le cas de gènes dont l’expression est souvent amenée à fluctuer sans passer par le remodelage de la chromatine et l’assemblage du complexe transcriptionnel. Ce type de régulation semble être utilisé dans le cas de BDF2 qui n’est pas régulé au niveau transcriptionnel (Roy et Chanfreau, 2014) ou de GLK1 pour lequel la répression transitoire causée par le passage en conditions anaérobiques est abolie en absence de RNT1 (chapitre 3, Figure supplémentaire 5). Finalement, dans le troisième modèle observé, Rnt1p réprime continuellement une fraction des ARNm produits de manière indépendante des conditions de cultures (Figure 9C). En maintenant des niveaux bas d’ARN, Rnt1p facilite la répression complète des transcrits en cas de besoin. Alternativement, il pourrait protéger l’expression génique des fluctuations de la transcription. Ce modèle serait utile dans le cas des gènes tel que RGT1, dont la régulation est principalement post-traductionnelle (Zaman et al., 2008). Naturellement, il est possible d’imaginer d’autres modèles de régulation ou alors des modèles hybrides. Il sera intéressant d’adapter des méthodes telles que le marquage au 4-thio-uridine, qui permettent de mesurer simultanément la cinétique 264

de production et de dégradation des ARNm (Rabani et al., 2011), afin de mieux comprendre la contribution de Rnt1p dans la répression conditionnelle des gènes. On peut toutefois conclure que Rnt1p régule ses substrats de nombreuses manières, en passant d’un clivage conditionnel à un clivage constitutif.

Figure 9 : Modèles hypothétiques de régulation conditionnelle de l’expression génique par Rnt1p. (A-C) Pour chaque modèle, l’image de gauche présente la situation dans laquelle l’expression d’un gène est active et celle du centre lorsque l’expression doit être réprimée. À droite, le graphique illustre la cinétique de répression de l’expression de chaque modèle en présence (RNT1; trait gris) ou en absence (rnt1∆; trait noir) de Rnt1p. Les ovales noirs et les « tartes » grises représentent la polymérase à ARN II (pol) et Rnt1p, respectivement.

265

7.3. Rnt1p cible les voies métaboliques sensibles aux conditions environnementales

La présente étude a permis d’identifier plusieurs nouveaux ARNm clivées in vitro par Rnt1p. Tel que vu aux chapitres III et V, plusieurs de ces cibles sont associées avec le métabolisme des hydrates de carbone (30 gènes) ainsi que la maintenance des mitochondries et la respiration (20 gènes). On retrouve aussi plusieurs gènes impliqués dans la réponse aux stress osmotique, chimique ou oxydatif (34 gènes), la biosynthèse des acides aminés (14 gènes), la réponse à l’inanition (12 gènes), ou la sécrétion et le transport vésiculaire (22 gènes). Il semble donc que Rnt1p clive principalement des gènes associés avec des voies métaboliques modulées en fonction des conditions de croissance de la cellule. En outre, la détection et le métabolisme des nutriments chez la levure sont étroitement liés avec la production des ribosomes, la croissance et la division cellulaire (Jorgensen et al., 2002; Jorgensen et Tyers, 2004; Zaman et al., 2008). Or, on sait déjà que Rnt1p est aussi impliqué dans la biogénèse des ribosomes et le cycle cellulaire (Abou Elela et al., 1996; Catala et al., 2004; Catala et al., 2008). La sécrétion est aussi en lien avec la synthèse des ribosomes via la réponse au stress des parois, un autre voie pour laquelle Rnt1p est impliqué (Catala et al., 2012; Li et al., 2000; Nierras et Warner, 1999). L’hypothèse générale qui découle de l’ensemble de ces observations et des résultats obtenus est que Rnt1p joue un rôle dans la coordination de la synthèse des ribosomes et de la division cellulaire en fonction des conditions environnementales. Il n’est toutefois pas clair si Rnt1p régule chacune de ces voies métaboliques de façon indépendante ou coordonnée. De plus, les mécanismes par lesquels les signaux environnementaux sont transmis et intégrés par Rnt1p afin de réguler le bon transcrit au bon moment restent encore à élucider.

Il est bien connu que la voie de signalisation TOR (mTOR chez l’humain) occupe un rôle central dans la coordination entre les signaux environnementaux, la réponse aux stress et la synthèse des ribosomes (Jorgensen et al., 2002; Jorgensen et Tyers, 2004; Zaman et al., 2008). En réponse à l’absence de 266

nutriments (glucose, azote), la voie TOR est notamment connue pour réguler l’initiation de la traduction, pour réprimer la transcription des gènes qui codent pour les protéines ribosomales, activer les voies de réponse aux stress, réguler le cycle cellulaire en plus d’induire la dégradation spécifique de certains ARNm (Albig et Decker, 2001; Cardenas et al., 1999; Shamji et al., 2000; Zaman et al., 2008). Ainsi, les voies métaboliques ciblées par la voie TOR sont très similaires à celles ciblées par Rnt1p. Par ailleurs, des résultats préliminaires suggèrent que l’exposition des cellules à la rapamycine (un inhibiteur de la voie TOR) réduit considérablement l’expression de l’ARNm de RNT1 et abolit la surexpression de l’ARNm IFH1 (un substrat de Rnt1p) observée dans la souche rnt1∆ (Mathieu Lavoie, données non montrées). En considérant ces observations, il serait intéressant d’approfondir l’hypothèse selon laquelle la voie de signalisation TOR régule Rnt1p, de manière directe ou indirecte, en réponse à la disponibilité des nutriments.

7.4. Rnt1p influence la transcription de certains gènes

Clairement, le nombre de gènes surexprimés in vivo en absence de RNT1 dépasse le nombre de cibles identifiées in vitro. Considérant l’impact pléiotropique de la délétion de RNT1 sur les phénotypes cellulaires, l’explication la plus logique est que l’effet sur l’expression de ces gènes résulte de l’adaptation de la cellule face à l’absence de la ribonucléase. Aussi, Rnt1p clive in vitro l’ARNm de plusieurs facteurs de transcription (MIG2, TUP1, SWI4, SNF2, etc) en plus d’interagir physiquement avec au moins 18 facteurs de transcription par essai de double hybride (Tremblay, 2002). Il est donc raisonable de croire que la délétion de RNT1 risque d’influencer, de manière directe ou indirecte, la régulation transcriptionnelle de nombreux gènes.

Il est indéniable que Rnt1p entretient des liens avec la transcription. En effet, Rnt1p interagit physiquement avec des sous-unités de l’ARN polymérase I, modifie le taux de synthèse des ARNr et participe à la terminaison de la transcription au locus ribosomal (Braglia et al., 2011; Catala et al., 2008). Des 267

essais de double-hybrides préliminaires ont également montré que Rnt1p peut interagir physiquement avec la queue C-terminale (CTD) de l'ARN polymérase II et participer à la sauvegarde de la terminaison de la transcription de certains ARNm (Ghazal et al., 2009). Parallèlement, des interactions physiques sont également démontrées entre Rnt1p et chacun des membres du complexe de terminaison de la transcription Nrd1p-Nab3p-Sen1p (Chinchilla et al., 2012; Gavin et al., 2006; Ursic et al., 2004; Vasiljeva et Buratowski, 2006). Bref, Rnt1p interagit physiquement et fonctionnellement avec la machinerie transcriptionnelle.

Tel que vu au chapitre V, la délétion de RNT1 affecte la production d’un gène rapporteur LacZ placé en aval des régions promotrices de MIG2, MTH1 et RGT1. Un résultat similaire a été obtenu avec la région promotrice du gène TLC1 (Larose, 2008). Puisque l’ARNm codant pour LacZ n’est pas un substrat de Rnt1p (Larose, 2008), il a été conclu que Rnt1p peut réprimer la transcription, indépendamment de la synthèse d’un ARN clivable par Rnt1p. Afin de mieux caractériser l’impact global de la délétion de RNT1 sur la transcription, des essais d’immunoprécipitation de la chromatine ciblant la sous-unité principale de l’ARN polymérase II (Rpb1p) couplés avec la détection par séquençage à haut-débit (ChIP-seq) ont été réalisés au laboratoire (Francis Malenfant et Mathieu Lavoie, données non publiées). Une analyse préliminaire des résultats obtenus n’a pas permis de confirmer que la délétion de RNT1 affecte de manière globale la transcription des gènes MIG2 ou TLC1. De son côté, l’expression de MTH1 et RGT1 est simplement trop faible dans les conditions de culture utilisées pour détecter l’association de l’ARN polymérase II par cette technique. Les profils de densité de lecture de séquençage pour ces gènes sont présentés en Annexe 1. En raison de ces résultats partiellement contradictoires, il sera nécessaire de réévaluer attentivement les résultats et les conclusions obtenus avec le gène rapporteur LacZ au chapitre V.

Malgré tout, une évaluation sommaire des résultats de l’essai ChIP-seq indique qu’environ 300 gènes montrent un profil d’association de l’ARN polymérase II significativement différent en absence de RNT1. Tant des gènes 268

codants et non-codants, clivés ou non par Rnt1p se retrouvent parmi cette liste. Par exemple, la délétion de Rnt1p semble provoquer une accumulation de l’ARN polymérase en aval de certains ARNno, comme snR47 dont le transcrit précurseur renferme un site de clivage de Rnt1p en 5’ (Annexe 1). Ce résultat suggère donc que Rnt1p affecte aussi la terminaison de la transcription de ces ARNno. Il est aussi intéressant de corréler cette liste de gènes avec les niveaux d’expression des ARN observés in vivo afin d’expliquer la surexpression de certains gènes dans la souche rnt1∆. Par exemple, l’association de l’ARN polymérase II au locus du gène NCW2 est grandement augmentée dans la souche rnt1∆ (Annexe 1). L’ARNm de NCW2 n’a pas été identifié comme substrat de Rnt1p malgré qu’il fait parti des transcrits les plus surexprimés dans la souche rnt1∆ (chapitre III). La protéine Ncw2p est une composante de la paroi cellulaire et son expression est induite en réponse à un stress des parois (Terashima et al., 2000). Puisque Rnt1p est directement impliqué dans la voie métabolique du stress des parois (Catala et al., 2012), NCW2 semble constituer un excellent exemple d’un gène dont l’expression est affectée de manière indirecte par la délétion de RNT1. En conclusion, Rnt1p affecte la transcription d’un ensemble distinct de gènes, soit de manière directe ou indirecte. Toutefois, les mécanismes requis pour cette régulation restent encore à élucider.

Parallèlement, un autre projet en cours au laboratoire vise à établir si le clivage à l’intérieur des transcrits peut amener la terminaison prématurée de la transcription. Ce projet partage donc des objectifs commun avec le travail présenté au chapitre IV, mais offre cette fois l’avantage d’analyser les substrats de Rnt1p dans leur contexte génétique naturel. Le clivage co-transcriptionnel couplé à la terminaison prématurée de la transcription serait évidemment un moyen très efficace de réprimer l’expression d’un gène.

269

7.5. La liaison de l’ARN indépendante du clivage laisse entrevoir un rôle non- catalytique de Rnt1p

Une étude publiée en 2004 a montré que l’expression d’un mutant catalytique de Rnt1p, qui est capable de lier l’ARN mais pas de le cliver, corrige certains des défauts du cycle cellulaire observés en absence de RNT1 (Catala et al., 2004). Ce résultat suggère donc que Rnt1p possède au moins une activité indépendante de la catalyse. Similairement, il a été observé chez la bactérie E. coli que la liaison de la RNase III permet de réguler la traduction de l’ARNm du gène cIII du bactériophage λ de manière indépendante du clivage (Altuvia et al., 1991). Étant donné l’abondance de structures tige-boucles NGNN dans le génome de S. cerevisiae (voir au chapitre III) et la capacité de Rnt1p à lier l’ARNdb de manière stable sans clivage (Lamontagne et Abou Elela, 2004), il est logique de croire que Rnt1p puisse réguler certains ARNm simplement en liant ceux-ci. La conséquence d’un tel mode de régulation serait, par exemple, la séquestration du transcrit au noyau de la cellule. De toute évidence, les techniques de détection des substrats développés au cours de ce projet ne permettent pas d’identifier les candidats qui seraient régulés de la sorte. Toutefois, des expériences additionnelles telles que des analyses de fluorescence par hybridation in situ (FISH) ou bien d’immunoprécipitation des ARNs liés à Rnt1p in vivo (ARN-ip) permettraient certainement de mieux comprendre le rôle de Rnt1p à titre de « protéine liant l’ARNdb ».

8. Les mécanismes de régulation de l’activité de Rnt1p

8.1. Les interactions protéines-protéines influencent le clivage de l’ARN par Rnt1p

Outre les mécanismes généraux de reconnaissance des substrats, nous ne possédons que peu d’informations sur la régulation de l’activité des RNase III en soi. Un des mécanismes probables de régulation des RNase III est certainement 270

les interactions protéines-protéines. Chez la bactérie, on sait que la proteine YmdB, dont l’expression est induite en conditions de stress, peut se lier au site actif de la RNase III. Cette interaction inactive la catalyse mais pas la liaison de l’ARNdb (Kim et al., 2008). Chez les eucaryotes, la protéine chaperonne DGCR8 est importante pour la sélection des substrats, mais aussi pour la stabilité de la protéine Drosha (Han et al., 2009). Chez Drosophila melanogaster, la maturation des pré-miARNs par Dicer 1 requiert l’association avec les protéines Loquacious qui possèdent trois dsRBD (Kim et al., 2009). Finalement, du côté de Rnt1p, il a été observé que son interaction avec la protéine Nop1p est requise pour le clivage de snR18 à un site non-canonique (Giorgi et al., 2001). En conclusion, l’interaction physique des RNAse III avec d’autres protéines cellulaires est susceptible d’influencer le clivage de l’ARN de différentes manières, soit en modulant l’activité catalytique, la sélection des substrats ou alors la stabilité de la protéine.

Précédemment, un cribblage double-hybride réalisé au laboratoire a permis d’identifier au moins 76 protéines pouvant interagir physiquement avec Rnt1p (Tremblay, 2002). De plus, une recherche dans les bases de données de Saccharomyces Genome Database (http://www.yeastgenome.org/) et de BioGrid (http://thebiogrid.org/) révèle qu’au moins 16 protéines additionnelles peuvent interagir physiquement avec Rnt1p. Malheureusement, plusieurs de ces 92 candidates sont des protéines cytoplasmiques qui ont été identifiées par des criblages à haut débit. Ainsi, il n’est pas clair quel pourcentage interagit réellement avec Rnt1p in vivo. À ce jour, l’interaction fonctionnelle avec Rnt1p a seulement été validé expérimentalement pour les protéines Nop1p, Gar1p, Nrd1p, Sen1p, Rpa12p, Rpa34p, Rpa49p (Catala et al., 2008; Giorgi et al., 2001; Tremblay, 2002; Vasiljeva et Buratowski, 2006). Néanmoins, le nombre élevé de protéines partenaires potentielles démontre le potentiel de Rnt1p pour former des interactions protéines-protéines.

Dans le but d’identifier de nouveaux substrats dépendants de la présence de protéines chaperonnes (tel qu’observé avec Nop1p sur snR18), j’ai employé la technique du « Cut and Chip » présentée au chapitre III sur des échantillons 271

d’ARN coupés par Rnt1p en présence d’extraits cellulaires totaux. Cet essai en présence d’ARN natif a notamment permis de confirmer que la majorité des cibles identifiés avec Rnt1p seul n’étaient pas le résultat d’un mauvais repliement de l’ARN découlant de la procédure d’extraction de l’ARN (chapitre III). En accord avec l’étude précédente (Giorgi et al., 2001), j’ai également pu confirmer que la présence d’extrait cellulaire favorise le clivage de snR18 dans son contexte naturel (Annexe 4). Plus intéressant encore, j’ai pu identifier ~63 substrats additionnels qui n’ont pas été détectés par les méthodes du « Cut and Chip » traditionnel ou de « SALI » présentés au chapitre III. Ces résultats suggèrent la présence de plusieurs substrats qui requièrent la participation de protéines chaperonnes pour être clivés par Rnt1p. Des essais supplémentaires seront toutefois requis pour valider cette affirmation.

8.2. Le recrutement co-transcriptionnel de Rnt1p

En 2011, Bregman et al. ont démontré que des éléments dans la région promotrice des gènes pouvaient affecter la dégradation cytoplasmique des transcrits qu’ils encodent (Bregman et al., 2011). De manière analogue, le remplacement de la région promotrice de deux substrats de Rnt1p, MIG2 et MTH1, par le promoteur du gène ACT1 a diminué l’impact sur l’expression observé lors de la délétion de RNT1 (chapitre V). Ainsi, l’identité de la région promotrice influence la régulation de l’ARN dépendante de Rnt1p. Considérant les liens entre Rnt1p et la transcription (voir la section 7.4), il a été proposé que Rnt1p s’associe avec le complexe transcriptionnel lors de l’initiation et/ou l’élongation de la transcription. En accord avec cette hypothèse, il a été montré que Rnt1p s’associe avec la chromatine au locus des gènes NPL3 et NAB2, deux substrats de Rnt1p (Ghazal et al., 2009). De plus, les auteurs de cette étude mentionnent que l’association de Rnt1p est dépendante de la transcription active. Afin de mieux caractériser l’association de Rnt1p avec l’ADN, des essais de ChIP-seq ont été réalisés en utilisant Rnt1p comme appât. L’analyse préliminaire des résultats montre que Rnt1p s’associe fortement avec la chromatine de plusieurs gènes clivés ou non par Rnt1p, notamment plusieurs snoRNAs (snR47, par exemple), la région 3'UTR de 272

NAB2 et la région intergénique NPL3-GPI17 (Annexe 2). On retrouve aussi quelques gènes qui ne sont pas connus pour être coupés par Rnt1p, tel que PDC1. Dans tous les cas, l’association est dépendante de l’ARN puisqu’aucune région ne semble enrichie dans les échantillons traités à la RNase A. À ce stade, il est difficile de confirmer si Rnt1p s’associe bel et bien avec les régions promotrices ou le corps des gènes MIG2, MTH1 ou RGT1 (Annexe 2). Bien qu’une analyse approfondie des données soit requise, les résultats préliminaires indiquent néanmoins que Rnt1p se retrouve à proximité de l’ADN de certains gènes spécifiques, en accord avec le modèle de recrutement co-transcriptionnel de Rnt1p.

Par ailleurs, l’examen sommaire de la liste des nouvelles protéines qui interagissent avec Rnt1p lors des cribblages révèle que plusieurs sont des facteurs de transcription (Tremblay, 2002). En outre, ceux-ci régulent l’expression de gènes impliqués dans des voies métaboliques liées au stress et souvent communes à celles régulées par Rnt1p. À titre d’exemple, Rnt1p interagirait avec les facteurs de transcription Gck2p et Adr1p (régulateurs des gènes de la glycolyse), Aro80p (régule la biogénèse des acides aminés), Rsf1p (régule les gènes mitochondriaux et impliqué dans la respiration) et Msn1p (réponse au choc osmotique). L’hypothèse qui découle de ces observations est que la liaison de facteurs de transcription facilite le recrutement de Rnt1p avec la machinerie transcriptionnelle.

En conclusion, de nombreuses évidences suggèrent que le clivage par Rnt1p se produit de manière co-transcriptionnel. En effet, Rnt1p influence la terminaison de la transcription, modifie l’association de l’ARN polymérase II sur la chromatine et est lui-même associé avec la chromatine de plusieurs de ses cibles d’une manière dépendante de l’ARN. Finalement, l’interaction avec des facteurs de transcription faciliterait le recrutement conditionnel de Rnt1p, assurant ainsi une régulation optimale.

273

8.3. La localisation subcellulaire de Rnt1p permettrait la régulation conditionnelle des gènes en fonction du cycle cellulaire

La localisation subcellulaire de Rnt1p pourrait être un autre moyen de réguler son activité. Les analyses de microscopie ont montré que Rnt1p est majoritairement localisé au noyau des cellules. Rnt1p ne semble pas présent au niveau du cytoplasme, mais les techniques employées ne permettent pas d’exclure formellement cette possibilité. De manière plus précise, Rnt1p est surtout observé au nucléole lors des phases G1 et S du cycle cellulaire et transige vers le nucléoplasme au cours des phases G2 et M (Catala et al., 2004). Ce patron de localisation est vraisemblablement lié au rôle de Rnt1p dans la maturation des pré- ARNr au niveau du nucléole (Catala et al., 2004; Henras et al., 2004). Puisque le passage de Rnt1p vers le nucléoplasme coïncide avec la diminution de l’ARN de EST1 (clivé par Rnt1p), il est proposé que la localisation de Rnt1p permet la régulation conditionnelle de EST1 en fonction du cycle cellulaire (Larose, 2008). Cette hypothèse reste toutefois à démontrer.

8.4. Les conditions de culture affectent le niveau d’expression de Rnt1p

Rnt1p est une protéine relativement abondante dans la cellule. On estime que chaque cellule contient en moyenne 1,4 ARNm du gène RNT1 et environ 4970 molécules de Rnt1p (Ghaemmaghami et al., 2003; Greenbaum et al., 2002). Le niveau d’ARNm de RNT1 est légèrement augmenté (1,7 à 2 fois) lors de l’induction du stress des parois (Catala et al., 2012) et légèrement réprimé (1,3 fois) lorsque les cellules se nourrissent de galactose au lieu de glucose (chapitre III). D’un autre côté, son expression est fortement réprimée (5 à 10 fois) lors d’un choc thermique (Catala et al., 2012) ou suite à l’ajout de rapamycine dans le milieu de culture (Mathieu Lavoie, données non publiées). Considérant que Rnt1p doit couper chacun des quelques 2000 pré-ARNr synthétisés à chaque minute (Warner, 1999), il est raisonnable de penser que seulement une fraction des molécules de Rnt1p est disponible pour la régulation des ARNm. Par conséquent, même une 274

diminution de 10% des niveaux de protéine pourrait potentiellement avoir un impact significatif sur la régulation des ARNm.

8.5. Rnt1p semble être régulé au niveau post-traductionnel

Les modifications post-traductionnelles constituent une autre façon de réguler l’activité des RNase III. Par exemple, la phosphorylation d’une sérine du RIIID de la RNase III chez E. coli par la protéine kinase du phage T7 stimule l’activité de la ribonucléase (Gone et Nicholson, 2012; Mayer et Schweiger, 1983). Chez Drosha, la phosphorylation de deux sérines de la région N-terminale est requise pour la localisation nucléaire de l’enzyme et la maturation des microARNs (Tang et al., 2010). Du côté de Rnt1p, un criblage à grande échelle des cibles de 87 kinases de la levure a révélé que 8 kinases (Dbf2p, Hsl1p, Ire1p, Ksp1p, Kss1p, Mck1p, Prk1p, Rck1p) peuvent phosphoryler Rnt1p, du moins in vitro. Il est intéressant de noter que Rnt1p clive l’ARNm de deux d’entre elles, soit Hsl1p et Ksp1p, ce qui laisse envisager un possible mécanisme d’autorégulation. Le logiciel de prédiction de sites de phosphorylation NetPhosYeast (http://www.cbs.dtu.dk/services/NetPhosYeast/) prédit 11 sites de phosphorylation dans la séquence de Rnt1p (Figure 10) (Ingrell et al., 2007). Au moins trois sites potentiels sont situés au niveau du signal de localisation nucléaire, donc susceptibles de réguler la localisation subcellulaire de Rnt1p. Certains sites de phosphorylation sont également susceptibles d’influencer la liaison et/ou la catalyse par Rnt1p. Les sérines 277 et 400, par exemple, sont situés respectivement à proximité d’un résidu important pour la formation du centre catalytique ou à l’extrémité du RBM 2 du dsRBD. La sérine 453 est particulièrement intéressante puisque ce résidu, situé au RBM 0, lie la guanosine conservée de la tétra-boucle et sa mutation diminue fortement le clivage de l’ARN (chapitre II). La phosphorylation d’aucun de ces sites n’a été validée expérimentalement in vitro ou in vivo. Malgré tout, les modifications post- transcriptionnelles de Rnt1p représentent une avenue intéressante pour étudier les signaux qui contrôlent la dégradation sélective de l’ARN. 275

Figure 10 : Sites potentiels de phosphorylation de Rnt1p. Capture d’écran montrant les résultats de l’analyse de la séquence protéique de Rnt1p par le logiciel NetPhosYeast (http://www.cbs.dtu.dk/services/NetPhosYeast/). Le logiciel calcule le potentiel de phosphorylation pour chaque sérine ou thréonine (traits verticaux). Les sites identifiés par leur lettre et leur position (exemple Serine 23 = S23) sont ceux qui sont considérés comme fortement probables par le logiciel (paramètres par défaut). Sous le graphique, le schéma représente la position relative des principaux domaines de Rnt1p (tels que décrits à la Figure 3), des motifs de liaison à l’ARN (RBMs; en rouge) et des résidus formant le cœur catalytique (étoiles).

9. À propos de la classification des RNase III

Historiquement, les membres de la famille des RNase III ont été classifiés en fonction de leurs domaines protéiques (voir la section 2.2. de l’introduction). Toutefois, la découverte continuelle de nouvelles RNase III ayant des structures intermédiaires rend cette classification quelque peu floue. Par exemple, les levures à bourgonnement Saccharomyces castellii, Candida albicans et Kluyveromyces polysporus possèdent des enzymes Dicer non canoniques qui n’ont pas de domaines PAZ ou hélicase (Drinnenberg et al., 2009). Dans le cas du Dicer de K. polysporus, la structure moléculaire ressemble davantage à celle de Rnt1p, mais son mode de reconnaissance des substrats s’apparente plus à celui de la RNase III de la bactérie Aquifex aeolicus (Weinberg et al., 2011). 276

Alternativement, on pourrait classifier les RNase III selon leur fonction cellulaire, comme par exemple, celles qui participent à la maturation des ARNs non-codants versus la production des siARNs versus le clivage des ARNm. Toutefois, cette classification n’est pas non plus idéale puisqu’on retrouve des RNase III multifonctionnelles, telle que l’enzyme Dicer de Candida albicans qui participe à la maturation des ARNs non codants et la production des siARNs (Bernstein et al., 2012).

On pourrait également classifier les RNase III selon leur mode de reconnaissance des substrats. Dans le cas de Rnt1p, on note plusieurs ressemblances avec celui de la RNase III bactérienne, telles que la conservation des RBMs 1 à 4 et la modulation de l'activité catalytique en fonction de la présence de paires de bases particulières à l'intérieur de l'ARNdb. D’un autre côté, Rnt1p montre aussi plusieurs similarités avec les enzymes Dicer canoniques, notamment dans l'implication de la portion N-terminale de l'enzyme dans la reconnaissance des substrats. En effet, Dicer utilise son domaine PAZ, et non pas le dsRBD pour lier ses substrats (MacRae et al., 2006). Dans les deux cas, la portion N-terminale influence le positionnement du site de clivage (Lavoie et Abou Elela, 2008; Liang et al., 2014; MacRae et Doudna, 2007; MacRae et al., 2006). Finalement, le choix du site de clivage est déterminé en fonction de la distance avec un motif structurel situé à l'extrémité, et non au centre, de l'hélice d'ARNdb.

Certains auteurs séparent la famille des RNase III en trois classes et incluent Rnt1p avec les RNase III bactériennes (Drider et Condon, 2004; MacRae et Doudna, 2007). Considérant l’ensemble des données recueillies au cours de cette thèse, il apparaît évident que Rnt1p est distinct des enzymes procaryotes, tant par sa structure, son mode de liaison des substrats et l’assemblage de son centre catalytique. Ainsi, la division de la famille des RNase III en quatre classes semble mieux appropriée. Toutefois, cette classification ne devrait pas reposer uniquement sur les domaines protéiques. Compte tenu de la diversité structurelle et fonctionnelle des différentes RNase III, elle doit aussi tenir compte des rôles 277

principaux joués par les différentes RNase III, leur mode de reconnaissance des substrats ainsi que des produits qu’elles génèrent. 278

CONCLUSION

Depuis quelques années, et particulièrement depuis la découverte du mécanisme d’inférence par l’ARN, il est devenu évident que la stabilité de l’ARN joue un rôle crucial dans la régulation de l’expression génique. Toutefois, les mécanismes qui dictent la dégradation sélective de l’ARN sont encore mal compris. Au cours de cette thèse, j’ai utilisé la ribonucléase III de S. cerevisiae comme modèle afin de comprendre comment des ARN sont sélectionnés pour la dégradation et d’évaluer sa contribution à la régulation de l’expression génique. Plus précisément, les travaux réalisés ont permis de mieux caractériser les déterminants chimiques et structuraux requis pour la reconnaissance des substrats, en plus d’identifier les gènes régulés par Rnt1p et d’évaluer les facteurs qui influencent la dégradation sélective des ARNm.

La modulation de la stabilité d’ARNm spécifiques par l’ARNi est aujourd’hui perçue comme une des approches thérapeutiques les plus prometteuses pour traiter diverses maladies cardiovasculaires, neurodégénérative ainsi que plusieurs types de cancers (Deng et al., 2014; Gandhi et al., 2014). Mais, contre toute attente, les mécanismes de sélection et de clivage de l’ARN pour deux des enzymes clés de l’ARNi, les RNAse III Dicer et Drosha, sont encore largement inconnus. Les données recueillies sur la structure de Rnt1p et sur son mécanisme de reconnaissance des substrats vont certainement aider à mieux comprendre le fonctionnement, la régulation et l’évolution des RNase III eucaryotes. Entre autre, les résultats obtenus avec Rnt1p ont permis d’observer pour la première fois la formation du complexe catalytique d’une RNase III eucaryote en plus d’expliquer comment l’enzyme lie un motif spécifique d’ARN et positionne le site de clivage. Bref, les travaux présentés dans cette thèse représentent une avancée significative dans la compréhension des mécanismes d’action des RNase III chez les organismes eucaryotes.

279

Parallèlement, les résultats ont également permis de mettre en lumière le rôle important de Rnt1p dans la régulation de l’expression des gènes. En effet, diverses approches génomiques ont démontré que le rôle cellulaire de Rnt1p ne se limite pas uniquement à la maturation des ARNs non-codants, mais aussi à la dégradation conditionnelle d’un grand nombre d’ARNm. Effectivement, Rnt1p coupe l’ARNm de nombreux gènes et régule leur expression d’une manière dépendante des conditions de cultures. Jusqu’à maintenant, les seuls mécanismes identifiés chez la levure capables de moduler la stabilité d’un ARNm normal se produisaient au cytoplasme, souvent en lien avec la traduction. Le fait que Rnt1p régule plusieurs ARNm à l’intérieur du noyau des cellules, parfois en association avec la transcription, démontre que le destin d’un ARN peut être décidé très tôt au cours de son cycle de vie. En conclusion, les résultats présentés dans cette thèse révèlent un nouveau mécanisme nucléaire de régulation conditionnelle de l’expression génique en fonction des conditions de culture.

Déjà, les connaissances acquises sur le mécanisme de reconnaissance des substrats de Rnt1p ont déjà permis le développement de plusieurs outils permettant la modulation de l’expression génique chez la levure (Babiskin et Smolke, 2011a, b, c; Lamontagne et Abou Elela, 2007; Lavoie et Abou Elela, 2008). Alternativement, il est aussi possible d’envisager la création de protéines recombinantes qui exploitent le mécanisme de reconnaissance particulier de Rnt1p afin de sélectionner des ARNs qui possède une tétra-boucle, comme par exemple l’ARN du génome de HIV (Stephenson et al., 2013). Finalement, il serait éventuellement possible d’adapter les méthodes de criblage biochimique développées ici à d’autres applications, comme par exemple identifier les cibles spécifiques et (surtout) non-spécifiques des petits ARNs interférants. En conclusion, ces connaissances acquises avec Rnt1p chez la levure peuvent mener à des applications concrètes et, on l’espère, faciliter le développement de nouveaux outils thérapeutiques.

Naturellement, plusieurs points restent toutefois à éclaircir, notamment sur les mécanismes qui permettent l’activation conditionnelle de la dégradation des 280

ARNs par Rnt1p. Comment Rnt1p détecte-t-il les changements environnementaux de la cellule et comment arrive-t-il à reconnaître le bon substrat, au bon moment? Plusieurs pistes ont été évoquées dans la discussion pour répondre à ces questions fondamentales. Les plus intéressantes à approfondir seront certainement les modifications post-traductionnelles ainsi que l’association de Rnt1p avec le complexe transcriptionnel.

281

REMERCIEMENTS

Je tiens d’abord à remercier mon directeur de thèse, le professeur Sherif Abou Elela, pour m’avoir accueilli dans son laboratoire, mais surtout pour m’avoir constamment encouragé à me surpasser. Je souhaite également remercier les professeurs Brendan Bell, Luc Gaudreau et Marvin Wickens qui ont gentiment accepté de participer à l’évaluation de cette thèse. Merci aussi aux équipes qui ont collaboré à mes travaux de recherche, notamment celles des professeurs Ambro van Hoof et Xinhua Ji. Bien entendu, je remercie grandement tous les membres passés et présents du laboratoire pour leur amitié, leur aide ainsi que les nombreuses discussions enrichissantes que nous avons eu. J’ai également une pensée pour tous mes collègues et amis qui ont contribué à rendre cette période de ma vie si agréable.

Un merci tout spécial à ma conjointe Dominique pour son support, sa compréhension et ses encouragements qui m’ont grandement aidé à traverser toutes ces années. Merci à mon fils Félix, une source de joie et de fierté sans pareil. Finalement, merci à ma famille pour leurs encouragements constants et leur soutien.

282

LISTE DES RÉFÉRENCES

Abou Elela, S., et Ares, M., Jr. (1998). Depletion of yeast RNase III blocks correct U2 3' end formation and results in polyadenylated but functional U2 snRNA. EMBO J 17(13), 3738-3746.

Abou Elela, S., Igel, H., et Ares, M., Jr. (1996). RNase III cleaves eukaryotic preribosomal RNA at a U3 snoRNP-dependent site. Cell 85(1), 115-124.

Alberts, B., Johnson, A., Lewis, J., Raff, M., Roberts, K., et Walter, P. (2002). Molecular Biology of the Cell, 4th Edition (New York, Garland Science).

Albig, A.R., et Decker, C.J. (2001). The target of rapamycin signaling pathway regulates mRNA turnover in the yeast Saccharomyces cerevisiae. Mol Biol Cell 12(11), 3428-3438.

Allmang, C., et Tollervey, D. (1998). The role of the 3' external transcribed spacer in yeast pre-rRNA processing. J Mol Biol 278(1), 67-78.

Altuvia, S., Kornitzer, D., Kobi, S., et Oppenheim, A.B. (1991). Functional and structural elements of the mRNA of the cIII gene of bacteriophage lambda. J Mol Biol 218(4), 723-733.

Ameres, S.L., Martinez, J., et Schroeder, R. (2007). Molecular basis for target RNA recognition and cleavage by human RISC. Cell 130(1), 101-112.

Babiskin, A.H., et Smolke, C.D. (2011a). Engineering ligand-responsive RNA controllers in yeast through the assembly of RNase III tuning modules. Nucleic Acids Res 39(12), 5299-5311.

Babiskin, A.H., et Smolke, C.D. (2011b). A synthetic library of RNA control modules for predictable tuning of gene expression in yeast. Mol Syst Biol 7471.

Babiskin, A.H., et Smolke, C.D. (2011c). Synthetic RNA modules for fine-tuning gene expression levels in yeast by modulating RNase III activity. Nucleic Acids Res 39(19), 8651-8664.

Badis, G., Saveanu, C., Fromont-Racine, M., et Jacquier, A. (2004). Targeted mRNA degradation by deadenylation-independent decapping. Mol Cell 15(1), 5-15.

Bernstein, D.A., Vyas, V.K., et Fink, G.R. (2012). Genes come and go: the evolutionarily plastic path of budding yeast RNase III enzymes. RNA Biol 9(9), 1123-1128.

283

Blaszczyk, J., Gan, J., Tropea, J.E., Court, D.L., Waugh, D.S., et Ji, X. (2004). Noncatalytic assembly of ribonuclease III with double-stranded RNA. Structure 12(3), 457-466.

Blaszczyk, J., Tropea, J.E., Bubunenko, M., Routzahn, K.M., Waugh, D.S., Court, D.L., et Ji, X. (2001). Crystallographic and modeling studies of RNase III suggest a mechanism for double-stranded RNA cleavage. Structure 9(12), 1225-1236.

Bond, U. (2006). Stressed out! Effects of environmental stress on mRNA metabolism. FEMS Yeast Res 6(2), 160-170.

Braglia, P., Kawauchi, J., et Proudfoot, N.J. (2011). Co-transcriptional RNA cleavage provides a failsafe termination mechanism for yeast RNA polymerase I. Nucleic Acids Res 39(4), 1439-1448.

Bregman, A., Avraham-Kelbert, M., Barkai, O., Duek, L., Guterman, A., et Choder, M. (2011). Promoter elements regulate cytoplasmic mRNA decay. Cell 147(7), 1473-1483.

Campbell, F.E., Jr., Cassano, A.G., Anderson, V.E., et Harris, M.E. (2002). Pre- steady-state and stopped-flow fluorescence analysis of Escherichia coli ribonuclease III: insights into mechanism and conformational changes associated with binding and catalysis. J Mol Biol 317(1), 21-40.

Cardenas, M.E., Cutler, N.S., Lorenz, M.C., Di Como, C.J., et Heitman, J. (1999). The TOR signaling cascade regulates gene expression in response to nutrients. Genes Dev 13(24), 3271-3279.

Catala, M., Aksouh, L., et Abou Elela, S. (2012). RNA-dependent regulation of the cell wall stress response. Nucleic Acids Res 40(15), 7507-7517.

Catala, M., Lamontagne, B., Larose, S., Ghazal, G., et Abou Elela, S. (2004). Cell cycle-dependent nuclear localization of yeast RNase III is required for efficient cell division. Mol Biol Cell 15(7), 3015-3030.

Catala, M., Tremblay, M., Samson, E., Conconi, A., et Abou Elela, S. (2008). Deletion of Rnt1p alters the proportion of open versus closed rRNA gene repeats in yeast. Mol Cell Biol 28(2), 619-629.

Chanfreau, G., Abou Elela, S., Ares, M., Jr., et Guthrie, C. (1997). Alternative 3'- end processing of U5 snRNA by RNase III. Genes Dev 11(20), 2741-2751.

Chanfreau, G., Buckle, M., et Jacquier, A. (2000). Recognition of a conserved class of RNA tetraloops by Saccharomyces cerevisiae RNase III. Proc Natl Acad Sci U S A 97(7), 3142-3147.

284

Chanfreau, G., Legrain, P., et Jacquier, A. (1998a). Yeast RNase III as a key processing enzyme in small nucleolar RNAs metabolism. J Mol Biol 284(4), 975- 988.

Chanfreau, G., Rotondo, G., Legrain, P., et Jacquier, A. (1998b). Processing of a dicistronic small nucleolar RNA precursor by the RNA endonuclease Rnt1. EMBO J 17(13), 3726-3737.

Chang, K.Y., et Ramos, A. (2005). The double-stranded RNA-binding motif, a versatile macromolecular docking platform. FEBS J 272(9), 2109-2117.

Chang, S.S., Zhang, Z., et Liu, Y. (2012). RNA interference pathways in fungi: mechanisms and functions. Annu Rev Microbiol 66305-323.

Chinchilla, K., Rodriguez-Molina, J.B., Ursic, D., Finkel, J.S., Ansari, A.Z., et Culbertson, M.R. (2012). Interactions of Sen1, Nrd1, and Nab3 with multiple phosphorylated forms of the Rpb1 C-terminal domain in Saccharomyces cerevisiae. Eukaryot Cell 11(4), 417-429.

Chong, M.M., Zhang, G., Cheloufi, S., Neubert, T.A., Hannon, G.J., et Littman, D.R. (2010). Canonical and alternate functions of the microRNA biogenesis machinery. Genes Dev 24(17), 1951-1960.

Conrad, C., et Rauhut, R. (2002). Ribonuclease III: new sense from nuisance. Int J Biochem Cell Biol 34(2), 116-129.

Conti, E., et Izaurralde, E. (2005). Nonsense-mediated mRNA decay: molecular insights and mechanistic variations across species. Curr Opin Cell Biol 17(3), 316- 325.

Correll, C.C., et Swinger, K. (2003). Common and distinctive features of GNRA tetraloops based on a GUAA tetraloop structure at 1.4 A resolution. RNA 9(3), 355- 363.

Court, D.L., Gan, J., Liang, Y.H., Shaw, G.X., Tropea, J.E., Costantino, N., Waugh, D.S., et Ji, X. (2013). RNase III: Genetics and function; structure and mechanism. Annu Rev Genet 47405-431.

Culbertson, M.R., et Neeno-Eckwall, E. (2005). Transcript selection and the recruitment of mRNA decay factors for NMD in Saccharomyces cerevisiae. RNA 11(9), 1333-1339.

Danin-Kreiselman, M., Lee, C.Y., et Chanfreau, G. (2003). RNAse III-mediated degradation of unspliced pre-mRNAs and lariat introns. Mol Cell 11(5), 1279-1289.

285

Davis, C.A., et Ares, M., Jr. (2006). Accumulation of unstable promoter-associated transcripts upon loss of the nuclear exosome subunit Rrp6p in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 103(9), 3262-3267.

Deng, Y., Wang, C.C., Choy, K.W., Du, Q., Chen, J., Wang, Q., Li, L., Chung, T.K., et Tang, T. (2014). Therapeutic potentials of gene silencing by RNA interference: principles, challenges, and new strategies. Gene 538(2), 217-227.

Dolken, L., Ruzsics, Z., Radle, B., Friedel, C.C., Zimmer, R., Mages, J., Hoffmann, R., Dickinson, P., Forster, T., Ghazal, P., et Koszinowski, U.H. (2008). High- resolution gene expression profiling for simultaneous kinetic parameter analysis of RNA synthesis and decay. RNA 14(9), 1959-1972.

Doyle, M., et Jantsch, M.F. (2002). New and old roles of the double-stranded RNA- binding domain. J Struct Biol 140(1-3), 147-153.

Drider, D., et Condon, C. (2004). The continuing story of endoribonuclease III. J Mol Microbiol Biotechnol 8(4), 195-200.

Drinnenberg, I.A., Fink, G.R., et Bartel, D.P. (2011). Compatibility with killer explains the rise of RNAi-deficient fungi. Science 333(6049), 1592.

Drinnenberg, I.A., Weinberg, D.E., Xie, K.T., Mower, J.P., Wolfe, K.H., Fink, G.R., et Bartel, D.P. (2009). RNAi in budding yeast. Science 326(5952), 544-550.

Egecioglu, D.E., Kawashima, T.R., et Chanfreau, G.F. (2012). Quality control of MATa1 splicing and exon skipping by nuclear RNA degradation. Nucleic Acids Res 40(4), 1787-1796.

Egli, M., et Gryaznov, S.M. (2000). Synthetic oligonucleotides as RNA mimetics: 2'- modified Rnas and N3'-->P5' phosphoramidates. Cell Mol Life Sci 57(10), 1440- 1456.

Eick, D., Piechaczyk, M., Henglein, B., Blanchard, J.M., Traub, B., Kofler, E., Wiest, S., Lenoir, G.M., et Bornkamm, G.W. (1985). Aberrant c-myc RNAs of Burkitt's lymphoma cells have longer half-lives. EMBO J 4(13B), 3717-3725.

Elkon, R., Zlotorynski, E., Zeller, K.I., et Agami, R. (2010). Major role for mRNA stability in shaping the kinetics of gene induction. BMC Genomics 11259.

Fasken, M.B., et Corbett, A.H. (2009). Mechanisms of nuclear mRNA quality control. RNA Biol 6(3), 237-241.

Filipowicz, W., Bhattacharyya, S.N., et Sonenberg, N. (2008). Mechanisms of post- transcriptional regulation by microRNAs: are the answers in sight? Nat Rev Genet 9(2), 102-114. 286

Filippov, V., Solovyev, V., Filippova, M., et Gill, S.S. (2000). A novel type of RNase III family proteins in eukaryotes. Gene 245(1), 213-221.

Findlay, D., Herries, D.G., Mathias, A.P., Rabin, B.R., et Ross, C.A. (1962). The active site and mechanism of action of bovine pancreatic ribonuclease. 7. The catalytic mechanism. Biochem J 85(1), 152-153.

Frischmeyer, P.A., et Dietz, H.C. (1999). Nonsense-mediated mRNA decay in health and disease. Hum Mol Genet 8(10), 1893-1900.

Gagnon, J., Lavoie, M., Catala, M., Malenfant, F., et Abou Elela, S. (2014). Transcriptome wide annotation of eukaryotic RNase III reactivity and degradation signals. Submitted for publication.

Gan, J., Shaw, G., Tropea, J.E., Waugh, D.S., Court, D.L., et Ji, X. (2008). A stepwise model for double-stranded RNA processing by ribonuclease III. Mol Microbiol 67(1), 143-154.

Gan, J., Tropea, J.E., Austin, B.P., Court, D.L., Waugh, D.S., et Ji, X. (2005). Intermediate states of ribonuclease III in complex with double-stranded RNA. Structure 13(10), 1435-1442.

Gan, J., Tropea, J.E., Austin, B.P., Court, D.L., Waugh, D.S., et Ji, X. (2006). Structural insight into the mechanism of double-stranded RNA processing by ribonuclease III. Cell 124(2), 355-366.

Gandhi, N.S., Tekade, R.K., et Chougule, M.B. (2014). Nanocarrier mediated delivery of siRNA/miRNA in combination with chemotherapeutic agents for cancer therapy: Current progress and advances. J Control Release 194C238-256.

Garneau, N.L., Wilusz, J., et Wilusz, C.J. (2007). The highways and byways of mRNA decay. Nat Rev Mol Cell Biol 8(2), 113-126.

Gaudin, C., Ghazal, G., Yoshizawa, S., Abou Elela, S., et Fourmy, D. (2006). Structure of an AAGU tetraloop and its contribution to substrate selection by yeast RNase III. J Mol Biol 363(2), 322-331.

Gavin, A.C., Aloy, P., Grandi, P., Krause, R., Boesche, M., Marzioch, M., Rau, C., Jensen, L.J., Bastuck, S., Dumpelfeld, B., Edelmann, A., Heurtier, M.A., Hoffman, V., Hoefert, C., Klein, K., Hudak, M., Michon, A.M., Schelder, M., Schirle, M., Remor, M., Rudi, T., Hooper, S., Bauer, A., Bouwmeester, T., Casari, G., Drewes, G., Neubauer, G., Rick, J.M., Kuster, B., Bork, P., Russell, R.B., et Superti-Furga, G. (2006). Proteome survey reveals modularity of the yeast cell machinery. Nature 440(7084), 631-636.

287

Ge, D., Lamontagne, B., et Elela, S.A. (2005). RNase III-mediated silencing of a glucose-dependent repressor in yeast. Curr Biol 15(2), 140-145.

Ghaemmaghami, S., Huh, W.K., Bower, K., Howson, R.W., Belle, A., Dephoure, N., O'Shea, E.K., et Weissman, J.S. (2003). Global analysis of protein expression in yeast. Nature 425(6959), 737-741.

Ghazal, G., et Abou Elela, S. (2006). Characterization of the reactivity determinants of a novel hairpin substrate of yeast RNase III. J Mol Biol 363(2), 332-344.

Ghazal, G., Gagnon, J., Jacques, P.E., Landry, J.R., Robert, F., et Abou Elela, S. (2009). Yeast RNase III triggers polyadenylation-independent transcription termination. Mol Cell 36(1), 99-109.

Ghazal, G., Ge, D., Gervais-Bird, J., Gagnon, J., et Abou Elela, S. (2005). Genome-wide prediction and analysis of yeast RNase III-dependent snoRNA processing signals. Mol Cell Biol 25(8), 2981-2994.

Giorgi, C., Fatica, A., Nagel, R., et Bozzoni, I. (2001). Release of U18 snoRNA from its host intron requires interaction of Nop1p with the Rnt1p endonuclease. EMBO J 20(23), 6856-6865.

Goldstrohm, A.C., Seay, D.J., Hook, B.A., et Wickens, M. (2007). PUF protein- mediated deadenylation is catalyzed by Ccr4p. J Biol Chem 282(1), 109-114.

Gone, S., et Nicholson, A.W. (2012). Bacteriophage T7 protein kinase: Site of inhibitory autophosphorylation, and use of dephosphorylated enzyme for efficient modification of protein in vitro. Protein Expr Purif 85(2), 218-223.

Greenbaum, D., Jansen, R., et Gerstein, M. (2002). Analysis of mRNA expression and protein abundance data: an approach for the comparison of the enrichment of features in the cellular population of proteins and transcripts. Bioinformatics 18(4), 585-596.

Guan, Q., Zheng, W., Tang, S., Liu, X., Zinkel, R.A., Tsui, K.W., Yandell, B.S., et Culbertson, M.R. (2006). Impact of nonsense-mediated mRNA decay on the global expression profile of budding yeast. PLoS Genet 2(11), e203.

Guthrie, C., et Fink, G.R. (1991). Guide to yeast genetics and molecular biology, Vol 194 (San Diego, Academic Press).

Han, J., Lee, Y., Yeom, K.H., Nam, J.W., Heo, I., Rhee, J.K., Sohn, S.Y., Cho, Y., Zhang, B.T., et Kim, V.N. (2006). Molecular basis for the recognition of primary microRNAs by the Drosha-DGCR8 complex. Cell 125(5), 887-901.

288

Han, J., Pedersen, J.S., Kwon, S.C., Belair, C.D., Kim, Y.K., Yeom, K.H., Yang, W.Y., Haussler, D., Blelloch, R., et Kim, V.N. (2009). Posttranscriptional crossregulation between Drosha and DGCR8. Cell 136(1), 75-84.

Hartman, E., Wang, Z., Zhang, Q., Roy, K., Chanfreau, G., et Feigon, J. (2013). Intrinsic dynamics of an extended hydrophobic core in the S. cerevisiae RNase III dsRBD contributes to recognition of specific RNA binding sites. J Mol Biol 425(3), 546-562.

Henras, A.K., Bertrand, E., et Chanfreau, G. (2004). A cotranscriptional model for 3'-end processing of the Saccharomyces cerevisiae pre-ribosomal RNA precursor. RNA 10(10), 1572-1585.

Houseley, J., et Tollervey, D. (2009). The many pathways of RNA degradation. Cell 136(4), 763-776.

Hu, W., Sweet, T.J., Chamnongpol, S., Baker, K.E., et Coller, J. (2009). Co- translational mRNA decay in Saccharomyces cerevisiae. Nature 461(7261), 225- 229.

Hughes, T.R., et de Boer, C.G. (2013). Mapping yeast transcriptional networks. Genetics 195(1), 9-36.

Iino, Y., Sugimoto, A., et Yamamoto, M. (1991). S. pombe pac1+, whose overexpression inhibits sexual development, encodes a ribonuclease III-like RNase. EMBO J 10(1), 221-226.

Ingrell, C.R., Miller, M.L., Jensen, O.N., et Blom, N. (2007). NetPhosYeast: prediction of protein phosphorylation sites in yeast. Bioinformatics 23(7), 895-897.

Ishigaki, Y., Li, X., Serin, G., et Maquat, L.E. (2001). Evidence for a pioneer round of mRNA translation: mRNAs subject to nonsense-mediated decay in mammalian cells are bound by CBP80 and CBP20. Cell 106(5), 607-617.

Johanson, T.M., Lew, A.M., et Chong, M.M. (2013). MicroRNA-independent roles of the RNase III enzymes Drosha and Dicer. Open Biol 3(10), 130144.

Jorgensen, P., Nishikawa, J.L., Breitkreutz, B.J., et Tyers, M. (2002). Systematic identification of pathways that couple cell growth and division in yeast. Science 297(5580), 395-400.

Jorgensen, P., et Tyers, M. (2004). How cells coordinate growth and division. Curr Biol 14(23), R1014-1027.

Kim, J. (2002). KEM1 is involved in filamentous growth of Saccharomyces cerevisiae. FEMS Microbiol Lett 216(1), 33-38. 289

Kim, K.S., Manasherob, R., et Cohen, S.N. (2008). YmdB: a stress-responsive ribonuclease-binding regulator of E. coli RNase III activity. Genes Dev 22(24), 3497-3508.

Kim, V.N., Han, J., et Siomi, M.C. (2009). Biogenesis of small RNAs in animals. Nat Rev Mol Cell Biol 10(2), 126-139.

Knight, S.W., et Bass, B.L. (2001). A role for the RNase III enzyme DCR-1 in RNA interference and germ line development in Caenorhabditis elegans. Science 293(5538), 2269-2271.

Kufel, J., Allmang, C., Chanfreau, G., Petfalski, E., Lafontaine, D.L., et Tollervey, D. (2000). Precursors to the U3 small nucleolar RNA lack small nucleolar RNP proteins but are stabilized by La binding. Mol Cell Biol 20(15), 5415-5424.

Kuperwasser, N., Brogna, S., Dower, K., et Rosbash, M. (2004). Nonsense- mediated decay does not occur within the yeast nucleus. RNA 10(12), 1907-1915.

Kuttykrishnan, S., Sabina, J., Langton, L.L., Johnston, M., et Brent, M.R. (2010). A quantitative model of glucose signaling in yeast reveals an incoherent feed forward loop leading to a specific, transient pulse of transcription. Proc Natl Acad Sci U S A 107(38), 16743-16748.

Lamontagne, B., et Abou Elela, S. (2001). Purification and characterization of Saccharomyces cerevisiae Rnt1p nuclease. Methods Enzymol 342159-167.

Lamontagne, B., et Abou Elela, S. (2004). Evaluation of the RNA determinants for bacterial and yeast RNase III binding and cleavage. J Biol Chem 279(3), 2231- 2241.

Lamontagne, B., et Abou Elela, S. (2007). Short RNA guides cleavage by eukaryotic RNase III. PLoS ONE 2(5), e472.

Lamontagne, B., Ghazal, G., Lebars, I., Yoshizawa, S., Fourmy, D., et Abou Elela, S. (2003). Sequence dependence of substrate recognition and cleavage by yeast RNase III. J Mol Biol 327(5), 985-1000.

Lamontagne, B., Hannoush, R.N., Damha, M.J., et Abou Elela, S. (2004). Molecular requirements for duplex recognition and cleavage by eukaryotic RNase III: discovery of an RNA-dependent DNA cleavage activity of yeast Rnt1p. J Mol Biol 338(2), 401-418.

Lamontagne, B., Larose, S., Boulanger, J., et Abou Elela, S. (2001). The RNase III family: a conserved structure and expanding functions in eukaryotic dsRNA metabolism. Curr Issues Mol Biol 3(4), 71-78. 290

Lamontagne, B., Tremblay, A., et Abou Elela, S. (2000). The N-terminal domain that distinguishes yeast from bacterial RNase III contains a dimerization signal required for efficient double-stranded RNA cleavage. Mol Cell Biol 20(4), 1104- 1115.

Larose, S. (2008). Étude des mécanismes de régulation de la télomérase chez la levure Saccharomyces cerevisiae. In Département de microbiologie et infectiologie (Sherbrooke, Université de Sherbrooke), pp. 245.

Larose, S., Laterreur, N., Ghazal, G., Gagnon, J., Wellinger, R.J., et Elela, S.A. (2007). RNase III-dependent regulation of yeast telomerase. J Biol Chem 282(7), 4373-4381.

Lau, P.W., Guiley, K.Z., De, N., Potter, C.S., Carragher, B., et MacRae, I.J. (2012). The molecular architecture of human Dicer. Nat Struct Mol Biol 19(4), 436-440.

Lavoie, M., et Abou Elela, S. (2008). Yeast ribonuclease III uses a network of multiple hydrogen bonds for RNA binding and cleavage. Biochemistry (Mosc) 47(33), 8514-8526.

Lavoie, M., Ge, D., et Abou Elela, S. (2011). Regulation of conditional gene expression by coupled transcription repression and RNA degradation. Nucleic Acids Res 40(2), 871-883.

Lebars, I., Lamontagne, B., Yoshizawa, S., Aboul-Elela, S., et Fourmy, D. (2001). Solution structure of conserved AGNN tetraloops: insights into Rnt1p RNA processing. EMBO J 20(24), 7250-7258.

Lee, A., Henras, A.K., et Chanfreau, G. (2005). Multiple RNA surveillance pathways limit aberrant expression of iron uptake mRNAs and prevent iron toxicity in S. cerevisiae. Mol Cell 19(1), 39-51.

Lee, C.Y., Lee, A., et Chanfreau, G. (2003). The roles of endonucleolytic cleavage and exonucleolytic digestion in the 5'-end processing of S. cerevisiae box C/D snoRNAs. RNA 9(11), 1362-1370.

Li, H., et Nicholson, A.W. (1996). Defining the enzyme binding domain of a ribonuclease III processing signal. Ethylation interference and hydroxyl radical footprinting using catalytically inactive RNase III mutants. EMBO J 15(6), 1421- 1433.

Li, W.M., Barnes, T., et Lee, C.H. (2010a). Endoribonucleases--enzymes gaining spotlight in mRNA metabolism. FEBS J 277(3), 627-641.

291

Li, Y., Liu, X., Huang, L., Guo, H., et Wang, X.J. (2010b). Potential coexistence of both bacterial and eukaryotic small RNA biogenesis and functional related protein homologs in Archaea. J Genet Genomics 37(8), 493-503.

Li, Y., Moir, R.D., Sethy-Coraci, I.K., Warner, J.R., et Willis, I.M. (2000). Repression of ribosome and tRNA synthesis in secretion-defective cells is signaled by a novel branch of the cell integrity pathway. Mol Cell Biol 20(11), 3843-3851.

Liang, Y.H., Lavoie, M., Comeau, M.A., Abou Elela, S., et Ji, X. (2014). Structure of a eukaryotic RNase III postcleavage complex reveals a double-ruler mechanism for substrate selection. Mol Cell 54(3), 431-444.

Libri, D., Dower, K., Boulay, J., Thomsen, R., Rosbash, M., et Jensen, T.H. (2002). Interactions between mRNA export commitment, 3'-end quality control, and nuclear degradation. Mol Cell Biol 22(23), 8254-8266.

Lim, B., Sim, S.H., Sim, M., Kim, K., Jeon, C.O., Lee, Y., Ha, N.C., et Lee, K. (2012). RNase III controls the degradation of corA mRNA in Escherichia coli. J Bacteriol 194(9), 2214-2220.

Lodish, H., Berk, A., Matsudaira, P., Kaiser, C.A., Krieger, M., Scott, M.P., Zipursky, L., et Darnell, J. (2003). Molecular cell biology, 5th Edition (New York, W.H. Freeman and Company).

MacRae, I.J., et Doudna, J.A. (2007). Ribonuclease revisited: structural insights into ribonuclease III family enzymes. Curr Opin Struct Biol 17(1), 138-145.

MacRae, I.J., Zhou, K., Li, F., Repic, A., Brooks, A.N., Cande, W.Z., Adams, P.D., et Doudna, J.A. (2006). Structural basis for double-stranded RNA processing by Dicer. Science 311(5758), 195-198.

Mayer, J.E., et Schweiger, M. (1983). RNase III is positively regulated by T7 protein kinase. J Biol Chem 258(9), 5340-5343.

Meaux, S., Lavoie, M., Gagnon, J., Abou Elela, S., et van Hoof, A. (2011). Reporter mRNAs cleaved by Rnt1p are exported and degraded in the cytoplasm. Nucleic Acids Res.

Meaux, S., et Van Hoof, A. (2006). Yeast transcripts cleaved by an internal ribozyme provide new insight into the role of the cap and poly(A) tail in translation and mRNA decay. RNA 12(7), 1323-1337.

Merritt, W.M., Lin, Y.G., Han, L.Y., Kamat, A.A., Spannuth, W.A., Schmandt, R., Urbauer, D., Pennacchio, L.A., Cheng, J.F., Nick, A.M., Deavers, M.T., Mourad- Zeidan, A., Wang, H., Mueller, P., Lenburg, M.E., Gray, J.W., Mok, S., Birrer, M.J., Lopez-Berestein, G., Coleman, R.L., Bar-Eli, M., et Sood, A.K. (2008). Dicer, 292

Drosha, and outcomes in patients with ovarian cancer. N Engl J Med 359(25), 2641-2650.

Michalova, E., Vojtesek, B., et Hrstka, R. (2013). Impaired pre-mRNA processing and altered architecture of 3' untranslated regions contribute to the development of human disorders. Int J Mol Sci 14(8), 15681-15694.

Moore, M.J. (2002). Nuclear RNA turnover. Cell 108(4), 431-434.

Munchel, S.E., Shultzaberger, R.K., Takizawa, N., et Weis, K. (2011). Dynamic profiling of mRNA turnover reveals gene-specific and system-wide regulation of mRNA decay. Mol Biol Cell 22(15), 2787-2795.

Nagel, R., et Ares, M., Jr. (2000). Substrate recognition by a eukaryotic RNase III: the double-stranded RNA-binding domain of Rnt1p selectively binds RNA containing a 5'-AGNN-3' tetraloop. RNA 6(8), 1142-1156.

Newbury, S.F. (2006). Control of mRNA stability in eukaryotes. Biochem Soc Trans 34(Pt 1), 30-34.

Nicholson, A.W. (2014). Ribonuclease III mechanisms of double-stranded RNA cleavage. Wiley Interdiscip Rev RNA 5(1), 31-48.

Nierras, C.R., et Warner, J.R. (1999). Protein kinase C enables the regulatory circuit that connects membrane synthesis to ribosome synthesis in Saccharomyces cerevisiae. J Biol Chem 274(19), 13235-13241.

Parker, R. (2012). RNA degradation in Saccharomyces cerevisae. Genetics 191(3), 671-702.

Qu, L.H., Henras, A., Lu, Y.J., Zhou, H., Zhou, W.X., Zhu, Y.Q., Zhao, J., Henry, Y., Caizergues-Ferrer, M., et Bachellerie, J.P. (1999). Seven novel methylation guide small nucleolar RNAs are processed from a common polycistronic transcript by Rat1p and RNase III in yeast. Mol Cell Biol 19(2), 1144-1158.

Rabani, M., Levin, J.Z., Fan, L., Adiconis, X., Raychowdhury, R., Garber, M., Gnirke, A., Nusbaum, C., Hacohen, N., Friedman, N., Amit, I., et Regev, A. (2011). Metabolic labeling of RNA uncovers principles of RNA production and degradation dynamics in mammalian cells. Nat Biotechnol 29(5), 436-442.

Raijmakers, R., Schilders, G., et Pruijn, G.J. (2004). The exosome, a molecular machine for controlled RNA degradation in both nucleus and cytoplasm. Eur J Cell Biol 83(5), 175-183.

Read, G.S. (2013). Virus-encoded endonucleases: expected and novel functions. Wiley Interdiscip Rev RNA 4(6), 693-708. 293

Robertson, H.D., Webster, R.E., et Zinder, N.D. (1968). Purification and properties of ribonuclease III from Escherichia coli. J Biol Chem 243(1), 82-91.

Romero-Santacreu, L., Moreno, J., Perez-Ortin, J.E., et Alepuz, P. (2009). Specific and global regulation of mRNA stability during osmotic stress in Saccharomyces cerevisiae. RNA 15(6), 1110-1120.

Rondon, A.G., Mischo, H.E., Kawauchi, J., et Proudfoot, N.J. (2009). Fail-safe transcriptional termination for protein-coding genes in S. cerevisiae. Mol Cell 36(1), 88-98.

Roy, K., et Chanfreau, G. (2014). Stress-induced nuclear RNA degradation pathways regulate yeast bromodomain factor 2 to promote cell survival. PLoS Genet 10(9), e1004661.

Schmid, M., et Jensen, T.H. (2008). Quality control of mRNP in the nucleus. Chromosoma 117(5), 419-429.

Schoenberg, D.R., et Maquat, L.E. (2012). Regulation of cytoplasmic mRNA decay. Nat Rev Genet 13(4), 246-259.

Seipelt, R.L., Zheng, B., Asuru, A., et Rymond, B.C. (1999). U1 snRNA is cleaved by RNase III and processed through an Sm site-dependent pathway. Nucleic Acids Res 27(2), 587-595.

Shalem, O., Dahan, O., Levo, M., Martinez, M.R., Furman, I., Segal, E., et Pilpel, Y. (2008). Transient transcriptional responses to stress are generated by opposing effects of mRNA production and degradation. Mol Syst Biol 4223.

Shamji, A.F., Kuruvilla, F.G., et Schreiber, S.L. (2000). Partitioning the transcriptional program induced by rapamycin among the effectors of the Tor proteins. Curr Biol 10(24), 1574-1581.

Sherman, F. (2002). Getting started with yeast. Methods Enzymol 3503-41.

Shyu, A.B., Wilkinson, M.F., et van Hoof, A. (2008). Messenger RNA regulation: to translate or to degrade. EMBO J 27(3), 471-481.

Stead, M.B., Marshburn, S., Mohanty, B.K., Mitra, J., Pena Castillo, L., Ray, D., van Bakel, H., Hughes, T.R., et Kushner, S.R. (2011). Analysis of Escherichia coli RNase E and RNase III activity in vivo using tiling microarrays. Nucleic Acids Res 39(8), 3188-3203.

294

Stephenson, J.D., Li, H., Kenyon, J.C., Symmons, M., Klenerman, D., et Lever, A.M. (2013). Three-dimensional RNA structure of the major HIV-1 packaging signal region. Structure 21(6), 951-962.

Sun, W., Jun, E., et Nicholson, A.W. (2001). Intrinsic double-stranded-RNA processing activity of Escherichia coli ribonuclease III lacking the dsRNA-binding domain. Biochemistry (Mosc) 40(49), 14976-14984.

Tang, X., Zhang, Y., Tucker, L., et Ramratnam, B. (2010). Phosphorylation of the RNase III enzyme Drosha at Serine300 or Serine302 is required for its nuclear localization. Nucleic Acids Res 38(19), 6610-6619.

Terashima, H., Yabuki, N., Arisawa, M., Hamada, K., et Kitada, K. (2000). Up- regulation of genes encoding glycosylphosphatidylinositol (GPI)-attached proteins in response to cell wall damage caused by disruption of FKS1 in Saccharomyces cerevisiae. Mol Gen Genet 264(1-2), 64-74.

Thapar, R., Denmon, A.P., et Nikonowicz, E.P. (2014). Recognition modes of RNA tetraloops and tetraloop-like motifs by RNA-binding proteins. Wiley Interdiscip Rev RNA 5(1), 49-67.

Thompson, D.M., et Parker, R. (2007). Cytoplasmic decay of intergenic transcripts in Saccharomyces cerevisiae. Mol Cell Biol 27(1), 92-101.

Trcek, T., Sato, H., Singer, R.H., et Maquat, L.E. (2013). Temporal and spatial characterization of nonsense-mediated mRNA decay. Genes Dev 27(5), 541-551.

Tremblay, A. (2002). Étude de la fonction de la RNase III eucaryote et identification de ses partenaires cellulaires dans un criblage double-hybrides. In Département de microbiologie et infectiologie (Sherbrooke, Université de Sherbrooke), pp. 230.

Tucker, M., Staples, R.R., Valencia-Sanchez, M.A., Muhlrad, D., et Parker, R. (2002). Ccr4p is the catalytic subunit of a Ccr4p/Pop2p/Notp mRNA deadenylase complex in Saccharomyces cerevisiae. EMBO J 21(6), 1427-1436.

Ursic, D., Chinchilla, K., Finkel, J.S., et Culbertson, M.R. (2004). Multiple protein/protein and protein/RNA interactions suggest roles for yeast DNA/RNA helicase Sen1p in transcription, transcription-coupled DNA repair and RNA processing. Nucleic Acids Res 32(8), 2441-2452. van Dijk, E.L., Chen, C.L., d'Aubenton-Carafa, Y., Gourvennec, S., Kwapisz, M., Roche, V., Bertrand, C., Silvain, M., Legoix-Ne, P., Loeillet, S., Nicolas, A., Thermes, C., et Morillon, A. (2011). XUTs are a class of Xrn1-sensitive antisense regulatory non-coding RNA in yeast. Nature 475(7354), 114-117.

295

Vasiljeva, L., et Buratowski, S. (2006). Nrd1 interacts with the nuclear exosome for 3' processing of RNA polymerase II transcripts. Mol Cell 21(2), 239-248.

Wang, Y., Liu, C.L., Storey, J.D., Tibshirani, R.J., Herschlag, D., et Brown, P.O. (2002). Precision and functional specificity in mRNA decay. Proc Natl Acad Sci U S A 99(9), 5860-5865.

Wang, Z., Hartman, E., Roy, K., Chanfreau, G., et Feigon, J. (2011). Structure of a yeast RNase III dsRBD complex with a noncanonical RNA substrate provides new insights into binding specificity of dsRBDs. Structure 19(7), 999-1010.

Warner, J.R. (1999). The economics of ribosome biosynthesis in yeast. Trends Biochem Sci 24(11), 437-440.

Weinberg, D.E., Nakanishi, K., Patel, D.J., et Bartel, D.P. (2011). The inside-out mechanism of Dicers from budding yeasts. Cell 146(2), 262-276.

Wu, H., Henras, A., Chanfreau, G., et Feigon, J. (2004). Structural basis for recognition of the AGNN tetraloop RNA fold by the double-stranded RNA-binding domain of Rnt1p RNase III. Proc Natl Acad Sci U S A 101(22), 8307-8312.

Wu, H., Yang, P.K., Butcher, S.E., Kang, S., Chanfreau, G., et Feigon, J. (2001). A novel family of RNA tetraloop structure forms the recognition site for Saccharomyces cerevisiae RNase III. EMBO J 20(24), 7240-7249.

Wyers, F., Rougemaille, M., Badis, G., Rousselle, J.C., Dufour, M.E., Boulay, J., Regnault, B., Devaux, F., Namane, A., Seraphin, B., Libri, D., et Jacquier, A. (2005). Cryptic pol II transcripts are degraded by a nuclear quality control pathway involving a new poly(A) polymerase. Cell 121(5), 725-737.

Yang, W. (2011). Nucleases: diversity of structure, function and mechanism. Q Rev Biophys 44(1), 1-93.

Yeom, K.H., Lee, Y., Han, J., Suh, M.R., et Kim, V.N. (2006). Characterization of DGCR8/Pasha, the essential cofactor for Drosha in primary miRNA processing. Nucleic Acids Res 34(16), 4622-4629.

Zaman, S., Lippman, S.I., Zhao, X., et Broach, J.R. (2008). How Saccharomyces responds to nutrients. Annu Rev Genet 4227-81.

Zenklusen, D., Vinciguerra, P., Wyss, J.C., et Stutz, F. (2002). Stable mRNP formation and export require cotranscriptional recruitment of the mRNA export factors Yra1p and Sub2p by Hpr1p. Mol Cell Biol 22(23), 8241-8253.

296

Zer, C., et Chanfreau, G. (2005). Regulation and surveillance of normal and 3'- extended forms of the yeast aci-reductone dioxygenase mRNA by RNase III cleavage and exonucleolytic degradation. J Biol Chem 280(32), 28997-29003.

Zhang, K., et Nicholson, A.W. (1997). Regulation of ribonuclease III processing by double-helical sequence antideterminants. Proc Natl Acad Sci U S A 94(25), 13437-13441.

297

ANNEXES

298

Annexe 1 : Résultats préliminaires de l’impact de la délétion de RNT1 sur l’ARN polymérase II.

Les graphiques montrent les densités de lectures de séquençage de six gènes (flèches grises) obtenues lors des expériences d’immunoprécipitation de la chromatine par l’ARN polymérase II. Brièvement, l’ARN polymérase II des souches de levures de type sauvage (gris) ou rnt1∆ (en noir) a été immunoprécipitée et l’ADN associé au complexe transcriptionnel quantifié par la technique de séquence à haut débit (ChIP-seq). L’échelle de l’axe des ordonnées (y) pour chaque graphique est indiquée entre les crochets. L’étoile présente la position relative du site de clivage de Rnt1p, s’il y a lieu. Voir la section 7.4. de la discussion pour plus de détails. La préparation des souches et l’immunoprécipitation a été réalisée par F. Malenfant. Les librairies de cDNA et l’analyse préliminaire des données de séquençage ont été faites par M. Lavoie.

299

Annexe 2 : Résultats préliminaires de l’association de Rnt1p avec la chromatine

Les graphiques montrent les densités de lectures de séquençage de sept gènes (flèches grises) obtenues lors des expériences d’immunoprécipitation de la chromatine par Rnt1p. Brièvement, Rnt1p a été immunoprécipité à partir d’échantillons traités (gris) ou non (noir) à la RNase A. L’ADN associé avec Rnt1p a ensuite été quantifié par la technique de séquence à haut débit (ChIP-seq). L’échelle de l’axe des ordonnées (y) pour chaque graphique est indiquée entre les crochets. L’étoile présente la position relative du site de clivage de Rnt1p, s’il y a lieu. Voir la section 8.2. de la discussion pour plus de détails. La préparation des souches et l’immunoprécipitation a été réalisée par F. Malenfant. Les librairies de cDNA et l’analyse préliminaire des données de séquençage ont été faites par M. Lavoie.

300

Annexe 3 : Liste des gènes clivés par Rnt1p in vitro (par ordre alphabétique).

Gènes Codants ARNs non-codants ACC1 DSE4 ISR1 PAN1 SEC26 TRS31 LSR1 snR57 ACF4 EAF1 IST2 PAN6 SEC3 TUB1 NME1 snR58 ACT1 EAR1 ITC1 PAT1 SEC31 TUP1 NTS1-2 snR59 ADI1 ECM29 IXR1 PEP1 SEC39 TYE7 RDN25-1 snR60 ADP1 ECM30 JJJ1 PEX25 SEC63 UBI4 SCR1 snR61 AHP1 EDE1 JSN1 PFF1 SEC7 UBP14 snR10 snR62 ALD5 EKI1 KCS1 PGA2 SFB3 UBP15 snR11 snR63 AMS1 ENO1 KIC1 PMA1 SFH5 URA2 snR14 snR64 APL6 ESC1 KIN2 PMC1 SIN3 URB1 snR161 snR65 ARN2 ESF2 KOG1 PMD1 SIN4 UTP14 snR17a snR66 ARO1 EST1 KRE1 PNC1 SKG3 UTP30 snR17b snR67 ASN1 FAS1 KRE33 PRP40 SLX8 VAC8 snR18 snR68 AVT1 FAS2 KSP1 PRP5 SMC4 VAS1 snR19 snR69 AXL2 FBA1 KTR5 PRP8 SMF1 VHT1 snR190 snR70 BAS1 FIP1 LIA1 PRP9 SML1 VMA1 snR3 snR71 BDF2 FIT2 MAP2 PRS5 SMY2 VPS13 snR30 snR72 BEM2 FKS1 MCM2 PSK1 SNA2 VPS72 snR36 snR73 BFR2 FMP27 MDL2 PSK2 SNF2 XBP1 snR39 snR74 BMH1 FRK1 MDN1 PTC2 SPC24 YAP1 snR39B snR75 BMS1 FUR1 MGA2 PTI1 SPE3 YAP1802 snR40 snR76 BNI1 GCD6 MGE1 PXR1 SPT5 YBP1 snR41 snR77 BSC1 GCN1 MIG2 RAT1 SRP101 YDL176W snR43 snR78 BSD2 GCV2 MIT1 RAX2 SRP72 YDR514C snR44 snR79 BTN2 GFA1 MLP1 RFC1 SSA2 YEF3 snR46 snR7-L BZZ1 GLK1 MLP2 RGT1 SSD1 YGL140C snR47 snR7-S CAF4 GLT1 MON2 RIF1 SSE1 YGR130C snR48 snR80 CCR4 GND1 MPP10 RLM1 SSE2 YHR080C snR50 snR81 CCT5 GPI16 MPT5 RPL10 SSY1 YIR007W snR51 snR83 CDC12 GPM1 MRPL51 RPL11A STE11 YKL162C snR52 snR84 CDC19 GPP1 MSB2 RPL15A STE6 YLR278C snR53 snR85 CEG1 GRE3 MTH1 RPL16B (i) STH1 YMR124W snR55 snR87 CHC1 GRS1 MTR3 RPL18A (i) SUI3 YNL115C snR56 TLC1 CHD1 GSH1 MYO1 RPL19A SVL3 YOR1

CMC2 (i) GYL1 NAB3 RPL19B SWH1 YPI1

CMK2 HAC1 NAM7 RPL27A (i) SWI1 YPL277C

COP1 HAS1 NAR1 RPL2B (i) SWI4 YPR117W

COQ1 HEH2 NCE102 RPL8A (3') SWR1 YSP2

COS111 HHT1 NHP2 RPN4 TAE2 YTA6

COT1 HMRA1 NOP14 RPO21 TAF6 ZRG8

CRH1 HRD1 NPL3-GPI17 RPO41 TAF7

CRN1 HSL1 NPR2 RPS13 TCB1

CTR9 HSP104 NUM1 RPS22B (i) TEF4

CWH41 HSP60 NUP159 RPS26B THO2 Autres transcrits CYS3 HTA1 NUP170 RPS31 TIF1 tA(AGC)J tQ(CUG)M YBRWTy1-2 DAM1 HTS1 NUP188 RRP14 TIM50 tA(UGC)A tQ(UUG)B YCLCdelta1 DBP2 (i) HUL5 NUP192 RRP5 TOM1 tA(UGC)G tQ(UUG)E2 YLRCTy2-2 DED1 HXK1 NUP60 RVS161 TOM71 tE(CUC)I tR(ACG)J YMRCTy1-3 DEF1 IFH1 OAR1 SAN1 TOP1 tE(UUC)M tR(CCG)L YNLCTy2-1 DGR2 ILS1 OMA1 SDA1 TOS2 tF(GAA)N tR(CCU)J YNLWTy1-2 DIS3 INP53 OSH6 SEC18 TPS1 tG(CCC)D tS(CGA)C YORWTy2-2 DNF3 IRA1 OSM1 SEC2 TRA1 tG(CCC)O tT(CGU)K chrV:503978- DSE1 IRA2 PAH1 SEC21 TRK1 tG(UCC)G YBLWTy2-1 505186 Les gènes préalablement identifiés comme substrats de Rnt1p sont indiqués en caractères gras. (i) et (3’) indiquent respectivement que le clivage a été détecté dans l’intron ou la région 3’UTR.

301

Annexe 4 : Essai de clivage par Rnt1p en présence d’extrait cellulaire total

Le clivage en présence d’extrait cellulaire permet l’identification de nouveaux substrats de Rnt1p. (A) La présence d’extrait cellulaire stimule le clivage de snR18 en fonction de l’augmentation de la concentration en KCl dans la réaction. L’ARN total ou l’extrait cellulaire respectivement préparés à partir de cellules drb1∆/rnt1∆ et dbr1∆ ont été incubés en absence ou présence de Rnt1p recombinant. L’ARN clivé a ensuite été analysé par essai d’extension d’amorce en utilisant une sonde radiomarquée complémentaire à l’intron du gène EFB1. La souche dbr1∆ a été utilisée afin de permettre l’accumulation du lariat. Les marqueurs de taille (M) sont indiqués à la gauche des gels. Un schéma du gène incluant le site de clivage de Rnt1p connu et les tailles attendues des ADN complémentaires est représenté au- dessus. N.S. indique un fragment non-spécifique détecté dans toutes les pistes. (B) La technique du Cut and Chip (Gagnon et al., 2014) a été appliquée à l’ARN natif clivé par Rnt1p en présence d’extrait cellulaire. Le diagramme de Venn illustre le nombre de genes codants (haut) ou d’ARNs non-codants (bas) identifiés comme clivés dans l’essai original (ARN purifié) ou en présence d’extrait cellulaire.