University of Massachusetts Medical School eScholarship@UMMS

GSBS Dissertations and Theses Graduate School of Biomedical Sciences

2020-07-22

Regulation of Translation and Synaptic Plasticity by TSC2

Annie Hien University of Massachusetts Medical School

Let us know how access to this document benefits ou.y Follow this and additional works at: https://escholarship.umassmed.edu/gsbs_diss

Part of the Molecular and Cellular Neuroscience Commons

Repository Citation Hien A. (2020). Regulation of Translation and Synaptic Plasticity by TSC2. GSBS Dissertations and Theses. https://doi.org/10.13028/zhjn-x778. Retrieved from https://escholarship.umassmed.edu/ gsbs_diss/1097

Creative Commons License

This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License This material is brought to you by eScholarship@UMMS. It has been accepted for inclusion in GSBS Dissertations and Theses by an authorized administrator of eScholarship@UMMS. For more information, please contact [email protected]. i

REGULATION OF TRANSLATION AND SYNAPTIC PLASTICITY BY TSC2

A Dissertation Presented

By

ANNIE HIEN

Submitted to the Faculty of the

University of Massachusetts Graduate School of Biomedical Sciences, Worcester

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

July 22th, 2020

Molecular Neuroscience

i ii

REGULATION OF TRANSLATION AND SYNAPTIC PLASTICITY BY TSC2

A Dissertation Presented

By

ANNIE HIEN

This work was undertaken in the Graduate School of Biomedical Sciences

M.D./Ph.D. Program Under the mentorship of

______Joel Richter, Ph.D., Thesis Advisor

______Daryl Bosco, Ph.D., Member of Committee

______Erik Sontheimer, Ph.D., Member of Committee

______Craig Peterson, Ph.D., Member of Committee

______Elizabeth Gavis, M.D./Ph.D., External Member of Committee

______Andrei Korostelev, Ph.D., Chair of Committee

______Mary Ellen Lane, Ph.D. Dean of the Graduate School of Biomedical Sciences

July 22th, 2020

ii iii

ACKNOWLEDGEMENTS

My thesis work would not be possible without the unwavering encouragement and guidance of my mentor, Dr. Joel Richter. Joel encourages me to think outside the box, explore my scientific curiosities, foster collaborations with others, and has improved my writing and vocabulary with phrases such as “not a big hairy deal”. Most importantly,

Joel taught me to be bold scientist (and person) with a voice. I am forever grateful for his dedication to my professional development. My thesis project would also not be possible without my excellent collaborators, Drs. Gemma Molinaro and Kimberly Huber from the

University of Texas Southwestern- this project took a heroic effort on everyone’s part.

Current and former members of the Richter lab have been invaluable for intellectual discussions, moral support, and mentorship- I could not ask for better colleagues from Drs. Lori Lorenz, Maria Ivshina, Huan (Raya) Shu, Sneha Shah, Heleen van't Spijker, Suna Jung, and fellow student Mohamad Nasrallah. In particular, Drs.

Botao Liu for contributing to my thesis project and teaching me ribosome profiling and the associated analysis, Emily Stackpole for teaching how to perform sucrose gradients and intellectual discussions related to FXR2P, Rhonda McFleder for being a mentor while I rotated in the lab and onward, and Ruijia (Ray) Wang for help with bioinformatics analysis.

I would like to thank members of my TRAC for their guidance and input over the years, Drs. Andrei Korostelev, Daryl Bosco, Erik Sontheimer, and Lihua (Julie) Zhu. I am also grateful for Dr. Craig Peterson at UMass and Dr. Elizabeth Gavis at Princeton

University for reviewing my thesis and serving on my Defense Exam Committee.

iii iv

I have been supported on this current career trajectory by the delicious home cooked meals and the enormous sacrifices made by my parents, Ying and Thuy, who immigrated to the USA without knowing English to provide better opportunities for my sisters and me. Through leading by example, you have instilled in me a diligent work ethics and resilience. For my community outside of lab, I want to thank my three sisters

Amy, Winnie, and Jessica, my friends at UMass, from Malden, the Kingdom Trail group, and climbing crew- you have all kept me sane and happy.

Finally, I want to acknowledge my favorite climbing, biking, and life partner,

Eric. I am grateful for your constant love and support through the sunny and cloudy days of the MD/PhD path. I look forward to our post-UMass adventures with our cats

(hopefully more to come).

iv v

ABSTRACT

Mutations in TSC2 cause the disorder tuberous sclerosis (TSC), which has a high incidence of autism and intellectual disability. TSC2 regulates mRNA translation required for group 1 metabotropic glutamate receptor-dependent synaptic long-term depression (mGluR-LTD), but the identity of mRNAs responsive to mGluR-LTD signaling in the normal and TSC brain is largely unknown. We generated Tsc2+/- mice to model TSC autism and performed ribosome profiling to identify differentially expressed genes following mGluR-LTD in the normal and Tsc2+/- hippocampus. Ribosome profiling reveals that in Tsc2+/- mice, RNA-binding targets of Fragile X Mental Retardation Protein

(FMRP) are increased. In wild-type hippocampus, induction of mGluR-LTD caused rapid changes in the steady state levels of hundreds of mRNAs, many of which are FMRP targets. Moreover, mGluR-LTD signaling failed to promote phosphorylation of eukaryotic elongation factor 2 (eEF2) in Tsc2+/- mice, and chemically mimicking phospho-eEF2 with low cycloheximide enhances mGluR-LTD in the Tsc2+/- brain. These results suggest a molecular basis for bidirectional regulation of synaptic plasticity by

TSC2 and FMRP. Furthermore, deficient mGluR-regulated translation elongation contributes to impaired synaptic plasticity in Tsc2+/- mice.

v vi

TABLE OF CONTENTS

Reviewer page ...... ii Acknowledgements ...... iii Abstract ...... v Table of Contents ...... vi List of Figures ...... ix List of Tables ...... xi Glossary of Terms ...... xii

Chapter I Introduction ...... 1

Overview of tuberous sclerosis ...... 1

Clinical features of tuberous sclerosis ...... 1

Genetics of tuberous sclerosis ...... 2

TSC is an inhibitor of mTOR signaling ...... 3

Neurological functions of TSC ...... 5

Treatment options for TSC ...... 6

Overview of Fragile X Syndrome ...... 7

Genetics of Fragile X Syndrome ...... 7

Clinical features of Fragile X Syndrome ...... 8

Molecular function of FMRP ...... 8

Treatment options for FXS ...... 9

Protein homeostasis in synaptic plasticity ...... 9

Optimal neural performance depends on protein homeostasis ...... 9

Some forms of memory require rapid protein synthesis ...... 10

Synaptic plasticity depends on local, activity-dependent translation in neurons ...... 12

Translational regulation in synaptic plasticity ...... 13 vi vii

Regulators of translation initiation in synaptic plasticity ...... 13

Regulation of translation elongation in synaptic plasticity ...... 21

Convergence of mGluR5 and mTOR signaling on ASD ...... 23

Chapter II Ribosome Profiling in Mouse Hippocampus: Plasticity-induced Regulation and Bidirectional Control by TSC2 and FMRP ...... 27

Preface ...... 28

Introduction ...... 28

Results ...... 32

Translational Profiling of the TSC hippocampus ...... 32

Dysregulation of 5′ terminal oligopyrimidine (TOP) mRNAs in Tsc2+/- ...... 37

FMRP binding targets are altered in opposite directions in TSC and FXS mice ...... 43

mGluR5 signaling induces a rapid RNA and translational response ...... 46

mGluR-induced translational dysregulation in TSC ...... 51

Signaling to eukaryotic elongation factor 2 is altered in the Tsc2+/- hippocampus ...... 53

Discussion ...... 59

Limitations ...... 67

Conclusions ...... 67

Materials and methods ...... 67

Acknowledgements ...... 77

Chapter III Global Profiling of Translation by FMRP-associated Ribosomes ...... 78

Preface ...... 79

Introduction ...... 79

Results ...... 84

FMRP interacts with ribosomal proteins after RNase digestion ...... 84

vii viii

Anti-FMRP enriches for specific RPFs over total ribosomes ...... 88

Anti-FMRP enriches for specific RPFs independent of FMRP ...... 93

SeRP in FMRP-FLAG neural progenitor cells ...... 97

Technical challenges to overcome for FMRP SeRP ...... 99

Discussion ...... 102

Overlapping function of FXRs and FMRP ...... 102

Auto-regulation of Fmr1 translation by FMRP ...... 103

FMRP-ribosome interaction ...... 104

Materials and methods ...... 107

Acknowledgements ...... 112

Chapter IV Discussion and Conclusions ...... 121

Summary ...... 121

FMRP binding targets connects mGluR-LTD with two mouse models of autism ...... 122

TSC2: potential regulator of mGluR-LTD signaling to eEF2 and FMRP ...... 123

Altered eEF2K/eEF2 signaling may underlie other phenotypes in Tsc2+/- mice ...... 124

Crosstalk of eEF2K/eEF2 signaling and TOP mRNAs in Tsc2+/- animals ...... 125

Role of mRNA turnover and stability in mGluR-LTD ...... 126

Limitations of mouse models of TSC and whole tissue sequencing ...... 129

Concluding remarks ...... 131

REFERENCES ...... 132

viii ix

LIST OF FIGURES

Figure I.1: Overview of TSC/mTOR signaling ...... 5

Figure I.2: Connection between synaptic plasticity and activity-dependent local

translation in the hippocampus ...... 11

Figure I.3: Major steps of cap-dependent translation initiation ...... 15

Figure I.4: Overview of translation signaling pathways important for mGluR-LTD ...... 20

Figure I.5: mGluR5 and mTORC1 signaling to translation ...... 26

Figure II.1: Schematic of experimental design ...... 33

Figure II.2: Ribosome profiling of Tsc2+/- hippocampal slices reveals RNA and RPF

dysregulation ...... 36

Figure II.3: Dysregulation of TOP mRNAs in Tsc2+/- ...... 38

Figure II.4: Lack of unique 5′UTR features in dysregulated RPFs from Tsc2+/- ...... 39

Figure II.5: Correlation of hippocampal TE with degree of dysregulation in Tsc2+/- ...... 41

Figure II.6: Transcript features of hippocampal TE ...... 42

Figure II.7: FMRP binding targets are regulated by TSC2 and FMRP in opposing

directions ...... 45

Figure II.8: mGluR signaling induces rapid RNA and RPF changes in wild-type

hippocampal slices ...... 48

Figure II.9: Gene set enrichment analysis of mGluR-responsive RNAs ...... 49

Figure II.10: Gene set enrichment analysis of mGluR-responsive RPFs ...... 50

Figure II.11: Some mRNAs altered in unstimulated Tsc2+/- slices do not respond to

DHPG-mGluR signaling ...... 52

ix x

Figure II.12: Deficient mGluR1/5-regulated signaling to translation elongation in Tsc2+/-

mice ...... 54

Figure II.13: Correlation of hippocampal TE with mGluR5-responsive mRNAs and

FMRP binding targets ...... 56

Figure II.14: Low dose cycloheximide enhances LTD in Tsc2+/- ...... 58

Figure II.15: Model of altered eEF2 signaling and FMRP binding targets in Tsc2+/- ...... 66

Figure III.1: Comparison of conventional and selective ribosome profiling ...... 83

Figure III.2: FMRP co-sediments with ribosomal proteins independent of RNA ...... 85

Figure III.3: FMRP interacts with ribosomal proteins independent of RNA ...... 87

Figure III.4: Gene body mapping frequency for conventional and selective ribosome

profiling ...... 89

Figure III.5: Anti-FMRP enriches for specific mRNAs ...... 90

Figure III.6: Different modes of binding by anti-FMRP ribosomes ...... 92

Figure III.7: Anti-FMRP enriches for specific mRNAs independent of FMRP ...... 94

Figure III.8: Anti-FMRP cross-reacts to paralogue FXR2P ...... 96

Figure III.9: FMRP-FLAG cosediments with polysomes similar to wild-type FMRP ..... 98

Figure III.10: Crosslinking stabilizes some FMRP-ribosome interaction ...... 101

x xi

LIST OF TABLES

Table III-1: List of sample types and purpose for FMRP selective ribosome profiling ... 88

Table III-2: List of 125 mRNAs enriched in anti-FMRP ribosomes ...... 113

xi xii

GLOSSARY OF TERMS

Full name Definition

A translation elongation inhibitor that is used to CHX cycloheximide immobilize ribosomes on mRNAs. A method that combines UV light cross-linking and Cross linking CLIP immunoprecipitation to identify mRNA targets of immunoprecipitation RNA-binding proteins. A statistically significant difference in expression levels between two experimental groups or DE differentially expressed conditions (knockout versus wild-type, treated versus untreated). A selective and potent agonist for group 1 metabotropic glutamate receptors 1 and 5 DHPG dihydroxyphenylglycine (mGluR1/5). Treatment with DHPG induces long- term depression in the hippocampus. A member of the GTP-binding elongation factor eukaryotic elongation eEF2 family that mediates translocation of the peptidyl- factor 2 tRNA from the A site to the P site of the ribosome. eukaryotic elongation A kinase that phosphorylates its sole substrate eEF2 eEF2K factor 2 kinase on Thr56 to inactivate the function of eEF2. An RNA-binding protein whose loss causes Fragile X Syndrome. FMRP binds mRNAs to control their FMRP Fragile X Related Protein transport, stability, and translation required for neuron-to-neuron communication. Fragile X Related A family of RNA-binding proteins that include FXR proteins FMRP and its paralogs FXR1P and FXR2P. The leading single-gene cause of intellectual FXS Fragile X Syndrome disability and autism spectrum disorder that arises from the absence of FMRP. A form of synaptic plasticity characterized by a LTD long-term depression persistent activity-dependent decrease in synaptic transmission that lasts greater than 1 hour. A form of synaptic plasticity characterized by a LTP long-term potentiation persistent activity-dependent increase in synaptic transmission that lasts greater than 1 hour. A form of LTD that depends on rapid protein metabotropic group 1 synthesis following activation of mGluR5 signaling mGluR- mediated long-term by DHPG. mGluR-LTD in this document refers LTD depression specifically to mGluR5-LTD in the CA1 region of the hippocampus. A G-protein-coupled receptor that is a member of the metabotropic glutamate group 1 metabotropic glutamate receptors. Activation mGluR5 receptor 5 of mGluR5 by the agonist DHPG leads to LTD in the hippocampus.

xii xiii

A multi-subunit serine/threonine kinase complex that mechanistic target of mTORC1 is a master regulator of cellular homeostasis and rapamycin complex 1 protein synthesis. A protein that binds RNA to regulate its processing, RBP RNA-binding protein localization, turnover, and translation A ~30nt portion of mRNA that is protected by the ribosome protected RPF ribosome after RNase digestion. RPFs reveal the fragment location and density of ribosomes on mRNA. A method to perform ribosome profiling on a specific Selective ribosome SeRP ribosome population, usually by enrichment from profiling affinity purification. A ratio of RPF to RNA levels commonly used in TE translation efficiency ribosome profiling experiments. A stretch of pyrimidines immediately after the 5′cap terminal oligopyrimidine of mRNAs. Present in many mRNAs that encode TOP tract components of the translation machinery and translation factors. An autosomal dominant disorder caused by loss of function mutations in TSC1 or TSC2. TSC can also tuberous sclerosis TSC refer to the protein complex made up of TSC1 and complex TSC2, which is a potent inhibitor of mTOR signaling. The protein product of TSC1, also known as tuberous sclerosis TSC1 hamartin. TSC1 interacts with TSC2 to stabilize the complex 1 complex TSC. The protein product of TSC2, also known as tuberin, tuberous sclerosis TSC2 which encodes a GTPase-activating protein. TSC2 complex 2 interacts with TSC1 to form the complex TSC. The regions upstream (5′) or downstream (3′) of the UTR untranslated region main coding sequence of an mRNA. Often contains regulatory sequences for RBPs to bind.

xiii 1

CHAPTER I INTRODUCTION

Overview of tuberous sclerosis

Clinical features of tuberous sclerosis Tuberous sclerosis (TSC) is a multi-organ disorder characterized by hamartomas, benign tumors that contain multiple cell types. Hamartomas in TSC are frequently observed in the skin, kidneys, lungs, heart, eyes, and brain. Because TSC1 and TSC2 encode tumor suppressors, individuals with TSC also have a higher risk of invasive malignancy. Roughly 80% of patients have hamartomatous lesions in the brain, referred to as cortical tubers. Cortical tubers form during early development and contain a mixture of undifferentiated, abnormal neurons and glia. Cortical tubers are generally benign and do not increase in size or number. Subependymal giant cell tumors (SEGA) are also commonly found in patients with TSC (~10%). SEGAs contain glial cells, and unlike cortical tubers, can slowly grow in size and number. In addition to the anatomical abnormalities, individuals with TSC also have a high incidence of neuropsychiatric symptoms that collectively form the tuberous sclerosis associated neuropsychiatric disorder (TAND) (Krueger and Northrup 2013). Individuals with TSC have a high incidence of epilepsy (90%), intellectual disability/cognitive deficits (~50%), autism

(~40%), and other behavioral problems such as aggression and attention deficit disorder.

The neurological symptoms have the largest burden on quality of life for individuals with

TSC and their caregivers (Hallett, Foster et al. 2011). Because the majority of patients with TSC have brain lesions, scientists and physicians initially proposed that the lesions

1 2

were the main driver for the neuropsychiatric symptoms. However, lesion burden is only mildly correlated with the severity of neurological complications. For example, seizures in TSC can originate outside of cortical hamartomatous legions. Furthermore, some patients with an anatomically normal brain can have severe cognitive dysfunction and conversely, patients with a high lesion load can have minimal cognitive impairment

(Smalley, Burger et al. 1994, Ridler, Suckling et al. 2004, Wong and Khong 2006).

Genetics of tuberous sclerosis TSC is an autosomal dominant disorder with an incidence of 1 in 6,000-10,000.

Roughly two-thirds of cases are caused by de novo mutations, and the remaining cases are familial (Northrup, Kwiatkowski et al. 1992, Sampson and Harris 1994). TSC is caused by loss of function of either tumor suppressor gene TSC1 or TSC2. Thousands of variants have been reported for TSC1/2 and include a wide spectrum of small deletions, insertions, missense, nonsense, and splicing site mutations (Rosset, Netto et al. 2017).

Most cases of TSC are sporadic; mutations in TSC2 are more common and associated with a more severe clinical presentation than mutations in TSC1 (Au, Williams et al.

2007). TSC is also a highly heterogeneous disorder with variable penetrance and expressivity. Some individuals with TSC remain undiagnosed for years, while others can have severe neurocognitive impairments requiring social services. Variable expressivity in TSC arises even within the same family, highlighting the role of other environmental and genetic modifiers (Smalley, Burger et al. 1994). Thus, genotype-phenotype correlations for TSC are challenging due to the thousands of reported mutations and variable expressivity.

2 3

TSC is an inhibitor of mTOR signaling TSC1 and TSC2 encode hamartin and tuberin respectively, which together form the protein complex tuberous sclerosis complex (TSC) (van Slegtenhorst, de Hoogt et al.

1997). TSC2/tuberin is a GTPase activating protein that inactivates the GTP-binding protein Ras homology enriched in brain (Rheb) (Sampson and Harris 1994, Inoki, Li et al. 2002). In response to upstream signaling from insulin, amino acids, oxygen levels, and growth factors that converge on AKT and ERK, TSC is inactivated by phosphorylation of

TSC2. Following inactivation of TSC, Rheb activates the mechanistic target of rapamycin complex 1 (mTORC1), a multi-unit serine/threonine kinase that is a master regulator of cell proliferation and growth (Inoki, Li et al. 2002, Inoki, Li et al. 2003,

Yang, Inoki et al. 2006). mTORC1 phosphorylates hundreds of downstream substrates to increase anabolic and decrease catabolic processes (Figure I.1).

mTORC1 broadly regulates cell homeostasis and growth through increasing translation, increasing specific gene expression programs, and inhibiting protein turnover.

Translational control is the best-studied mTORC1-dependent process. mTORC1 signaling broadly increases protein synthesis capacity of cells by directly activating translation factors and regulators. In addition, mTORC1 signaling increases translation of mRNAs encoding the translation machinery and translation factors to further support increased protein synthesis. In parallel with the translational response, mTORC1 can activate the following transcription factors to support anabolic processes: lipogenesis via

SREBP (Düvel, Yecies et al. 2010), glycolysis via HIF1a (Hudson, Liu et al. 2002,

Düvel, Yecies et al. 2010), and mitochondria biogenesis via PGC1a. mTORC1 also

3 4

directly reduces protein turnover by inhibiting autophagy, lysosome biogenesis, and proteasome assembly (Peña-Llopis, Vega-Rubin-de-Celis et al. 2011, Settembre, Zoncu et al. 2012, Di Nardo, Wertz et al. 2014).

TSC-Rheb-mTORC1 is the canonical signaling pathway, but TSC can regulate non-mTORC1 processes (Neuman and Henske 2011). TSC controls cytoskeletal dynamics and cell adhesion by directly binding to the ezrin-radixin-moesin family of actin-binding proteins and activating the small GTPase Rho (Lamb, Roy et al. 2000,

Astrinidis, Cash et al. 2002). Rheb can also activate processes independent of mTORC1.

Overexpression of inactivated Rheb rescues dendritic spine morphogenesis in TSC- deficient neurons, but not mTORC1 knockdown or rapamycin treatment, suggesting that this phenotype is mediated by Rheb hyperactivation independent from mTORC1 signaling. Finally, there is also cross talk between mTORC1 and mTORC2, a structurally and functionally distinct mTOR complex with roles in cytoskeletal rearrangement. TSC is required for proper mTORC2 function through AKT signaling (Huang and Manning

2009).

4 5

Figure I.1: Overview of TSC/mTOR signaling mTOR integrates signaling from multiple membrane receptors to maintain cellular homeostasis. The best characterized mTOR-dependent processes are depicted. TrkB: tyrosine receptor kinase B, NMDA-R: N-methyl-D- aspartate receptor, mGluR5: metabotropic glutamate receptor 5. Accessory proteins for mTORC1 and mTORC2 are excluded for simplicity. Colors indicate positive regulators (green) and negative regulators (red) of mTOR signaling.

5 6

Neurological functions of TSC TSC/mTOR signaling has a role in maintenance of neural stem cells, neurite growth, and synaptic plasticity. Proper mTOR signaling is required for neural stem cell proliferation and differentiation. TSC2-deficient neural stem cells exhibit increased proliferation and favor glial over neuronal differentiation, which may contribute to the development of cortical tubers during embryogenesis (Grabole, Zhang et al. 2016). TSC is also required for proper dendritic and axonal growth. In Tsc1 haploinsufficient mice, neurons have increased soma size and decreased spine density (Tavazoie, Alvarez et al.

2005). Tsc2-deficient neurons have increased number of axons via increased activity of

SAD kinase, an important regulator of neuron polarity and axonogenesis (Choi, Di Nardo et al. 2008). TSC2 also regulates axon guidance in the visual system in an ephrin-Eph dependent manner (Nie, Di Nardo et al. 2010).

Treatment options for TSC Because excessive mTORC1 signaling underlies many pathogenic mechanisms arising from TSC haploinsufficiency, chemical inhibitors of mTOR are the main TSC- specific therapeutic (Hallett, Foster et al. 2011, Sadowski, Kotulska et al. 2016).

Everolimus is FDA-approved for treatment of several TSC-specific pathologies.

Everolimus has been shown to decrease tumor size in SEGA and is used in patients who are non-surgical candidates (Krueger, Care et al. 2010). Roughly two-thirds of patients with TSC have seizures refractory to standard anti-epileptic drugs. In these cases, everolimus is effective in treating the refractory seizures by reducing frequency in 90%

6 7

and completely eliminating seizures in 33% of subjects (French, Lawson et al. 2016). For the neuropsychiatric symptoms of TSC, there currently are no TSC-specific treatments.

Several randomized control clinical trials found that everolimus did not improve neurocognition, intellectual disability, or autism in children with TSC (Krueger,

Sadhwani et al. 2017, Overwater, Rietman et al. 2019). This was a disappointing result in the field, despite evidence in mouse models of TSC that demonstrated improvement in various behavioral assays with rapamycin. Thus, the therapeutic options for the treatment of neuropsychiatric symptoms of TSC remain limited (Hallett, Foster et al. 2011). A remaining question is if other rapamycin-insensitive processes contributes to ASD-like behaviors in humans and mouse models.

Overview of Fragile X Syndrome

Genetics of Fragile X Syndrome Fragile X Syndrome (FXS) is the leading single-gene cause of intellectual disability and roughly one-third of individuals also have autism spectrum disorder (ASD).

Initial karyotypes of FXS patients showed a piece missing from the X chromosome, from which the disorder is named (Oberlé, Rousseau et al. 1991). FXS is an X-linked disorder caused by a CGG repeat expansion in the 5′ untranslated region (5′UTR) of the Fragile X

Mental Retardation 1 gene (FMR1) (Verkerk, Pieretti et al. 1991). Normal individuals have <50 CGG repeats and permutation carriers have 50 to 200 repeats. With over 200 repeats, FMR1 is methylated and expression of the protein product Fragile X Mental

Retardation Protein (FMRP) is reduced or absent, leading to the disorder (Pieretti, Zhang et al. 1991). The incidence of FXS ranges from 1 in 4,000-7,000 in males. Females can

7 8

also be affected, though usually in a milder manner. Although premutation carriers do not develop FXS, they can present with Fragile X-associated tremor/ataxia, a late onset neurodegenerative disorder with neurocognitive impairment (Bagni, Tassone et al. 2012).

Clinical features of Fragile X Syndrome The main symptoms of FXS are developmental delays, learning disabilities, and behavioral problems. Many individuals have intellectual disability and language delays, which can range from mild to severe. Over half of males with FXS have ASD, and FXS is one of the leading genetic syndromes associated with ASD (Richards, Jones et al.

2015). Other common features include sensory disorders, anxiety, attention deficit disorder, aggression, and seizures (Bagni, Tassone et al. 2012). Physical characteristics of males with FXS include an elongated face, large ears, a prominent jaw, and macroorchidism. Infants with FXS can have macrocephaly, but otherwise display no brain structural abnormalities.

Molecular function of FMRP FMR1 encodes Fragile X Mental Retardation Protein (FMRP), an RNA binding protein required for normal brain development and synaptic function. In mice, FMRP binds roughly 5% of brain mRNAs to regulate mRNA transport, stability, and translational repression of its targets. (Darnell, Van Driesche et al. 2011, Maurin,

Lebrigand et al. 2018, Sawicka, Hale et al. 2019). Targets of FMRP overlap with genes linked to ASD and synaptic structure and function, and an important role of FMRP is to tightly control the synaptic proteome in response to neuronal activity (Huber, Gallagher et al. 2002, Bagni and Zukin 2019).

8 9

Treatment options for FXS There currently is no FXS-specific treatment or cure. For FXS and other developmental disorders such as TSC and ASD, a treatment plan involves a multi- disciplinary approach with early intervention services and medication to address behavioral problems. FDA-approved drugs for the management of behavioral problems in FXS and ASD are anti-psychotics for aggression and stimulants for attention deficit disorders. Based on promising preclinical results in mice, metabotropic glutamate receptor 5 (mGluR5) inhibitors were developed to treat FXS (Bear, Huber et al. 2004,

Dölen, Osterweil et al. 2007). However, randomized control studies using the mGluR5 inhibitor mavoglurant did not improve behavioral symptoms in adolescents and adults with FXS (Berry-Kravis, Des Portes et al. 2016). Because a criticism of these clinical trials included the use of subjects who are past the optimal window of brain plasticity, mGluR5 inhibitors are still being evaluated as a therapeutic option for FXS.

Protein homeostasis in synaptic plasticity

Optimal neural performance depends on protein homeostasis Certain brain regions exhibit high levels of plasticity during neurodevelopment, where neural circuits have a heighted sensitivity to inputs. Because ASD is a neurodevelopmental disorder, disruptions during this critical period of plasticity have a large and lasting impact on neural circuit maturation. Kelleher and Bear proposed that optimal neuronal performance depends on tightly regulated synaptic protein synthesis, and when synaptic protein homeostasis is disrupted, the subsequently altered neural circuitry can contribute to ASD (Kelleher and Bear 2008). Support for their model

9 10

comes from the observation that 1) multiple genetic syndromes that have a high prevalence of ASD encode for translation repressors (e.g. FMRP in Fragile X Syndrome,

TSC1/2 in TSC-associated neuropsychiatric disorder, and PTEN-associated autism), 2) multiple mouse models of ASD have altered global protein synthesis in the brain, and 3) restoring protein homeostasis in some mouse models of ASD can rescue molecular, synaptic, and behavioral phenotypes associated with autism (Auerbach, Osterweil et al.

2011, Bhattacharya, Kaphzan et al. 2012, Udagawa, Farny et al. 2013).

Some forms of memory require rapid protein synthesis Rapid protein synthesis in the hippocampus is required for some forms of long- term memories (Hernandez and Abel 2008). Flexner et al first demonstrated that application of puromycin (a premature nascent chain terminator of protein synthesis) prevents mice from learning a discriminative avoidance task in mice (Flexner, Flexner et al. 1962, Flexner, Flexner et al. 1963). Thus, how is the requirement for protein synthesis in some forms of learning and memory translated on a molecular level? Synaptic plasticity is the ability of synapses to change persistently following a stimulus and is believed to underlie learning and memory. There are two forms of synaptic plasticity: long-term potentiation (LTP) and long-term depression (LTD). LTP is a persistent strengthening of a synaptic connection whereas LTD is a persistent weakening. Synaptic plasticity is best studied at the Schaffer collateral synapses in the CA1 region of hippocampus, depicted in Figure I.2. In this region, two forms of synaptic plasticity that depend on rapid dendritic protein synthesis are late-LTP (L-LTP) and metabotropic glutamate receptor group 1 (mGluR) LTD. mGluR-LTD is mediated by activation of

10 11

mGluR1/5 (G-protein-coupled receptors) with the agonist dihydroxyphenlglycine

(DHPG) (Huber, Kayser et al. 2000).

Figure I.2: Connection between synaptic plasticity and activity-dependent local translation in the hippocampus

Left: The organization of the hippocampus. The Schaffer-collateral pathway is composed of axons from excitatory pyramidal neurons in the CA3 region (blue) that synapse on pyramidal neurons in CA1 (red). For long-term depression (LTD) experiments, the Schaffer-collateral pathway is stimulated and field potentials are recorded in the stratum radiatum (s. radiatum) of CA1, an area enriched for dendrites of CA1 pyramidal neurons. Middle: Treatment of hippocampal slices with dihydroxyphenlglycine (DHPG) induces LTD. The dashed line indicates the field excitatory postsynaptic potential (fEPSP) slope baseline before DHPG. The fEPSP slope is measured by a recording electrode placed in the s. radiatum of CA1 and reflects the excitatory response of the CA1 neurons from the CA3 neurons. After a brief application with DHPG, the fEPSP slope is persistently weaker and lasts for greater than 60 minutes. LTD is the difference between the pre-DHPG and post- DHPG fEPSP. Right: Rapid, local translation in the dendrites of neurons is required for mGluR-LTD.

11 12

Synaptic plasticity depends on local, activity-dependent translation in neurons Neurons are highly polarized cells. Most neurons contain a cell body (soma), 1 axon, and multiple, branched dendrites. Collectively, the neuronal processes make up the majority of the cytoplasm and are quite long compared to the soma (Holt, Martin et al.

2019). In vertebrates, axons can extend hundreds of centimeters, and the average dendrite length of a pyramidal neuron is 13.5mm (Ishizuka, Cowan et al. 1995). Neuronal communication occurs at the synapse, where neurotransmitters are released from the presynaptic axon to activate receptors on the postsynaptic membrane. An excitatory pyramidal neuron receives input from roughly 54,000 synapses from different neurons, and the highly polarized nature of neurons poses a unique challenge in how neurons can rapidly and selectively respond to a specific, local stimulus (Holt, Martin et al. 2019).

One such mechanism is through local protein synthesis that occurs in dendrites.

Rapid protein synthesis in dendrites of neurons is required for mGluR-LTD and some forms of LTP. Kang and Schuman demonstrated that LTP induced by neurotrophic factors BDNF and NT-3 requires rapid protein synthesis; co-application with translation inhibitors anisomycin or cycloheximide blocked these forms of LTP (Kang and Schuman

1996). To demonstrate a role for dendritic protein synthesis, Kang and Schuman micro- dissected the cell body from the dendrites of the CA1 pyramidal neurons and still observed induction of LTP that was dependent on protein synthesis. Using similar experiments, Huber et al demonstrated that DHPG induces LTD (mGluR-LTD) in the same region of the hippocampus, and mGluR-LTD also requires rapid de novo protein synthesis in dendrites (Huber, Kayser et al. 2000).

12 13

The newly synthesized proteins required for synaptic plasticity are termed LTP or

LTD proteins, and they underlie some of the local remodeling required for synaptic plasticity (Lüscher and Huber 2010). For example, microtubule-associated protein 1b

(MAP1B) and activity-regulated cytoskeleton-associated protein (ARC) are synthesized in dendrites following mGluR-LTD, and both function to decrease AMPA receptor trafficking to postsynaptic membranes (Davidkova and Carroll 2007, Park, Park et al.

2008, Waung, Pfeiffer et al. 2008). Dendritic translation of CaMKII is required for LTP, learning and memory, as demonstrated in a mouse model with impaired dendritic localization of CaMKII mRNA. In this mouse model, the dendrite localization signal of

CaMKII mRNA was mutated, and this mouse had reduced LTP and had deficits in several protein-synthesis dependent behavioral tasks (spatial memory, fear conditioning, and object placement) (Miller, Yasuda et al. 2002). The identity of the plasticity proteins required for mGluR-LTD is largely unknown and remains a question of great interest.

Recent advances in unbiased, deep-sequencing methods will allow for identification of these plasticity proteins.

Translational regulation in synaptic plasticity

Regulators of translation initiation in synaptic plasticity Translation is the process of producing proteins from mRNA by ribosomes.

Because multiple inhibitors of the ribosome prevent synthesis of proteins required for synaptic plasticity, research has focused on translational regulation. Translational occurs broadly in 3 steps: initiation, elongation, and termination. Translation initiation is the process of the small ribosomal subunit binding to the mRNA and recognizing the

13 14

appropriate initiation codon. Translation initiation is the most complex step, involves multiple translation factors, and often considered the rate-limiting step of translation. The first step of initiation is the formation of the 43S preinitiation complex from the 40S and eIF2 ternary complex. The 43S complex then attaches to the mRNA in a 5′cap-dependent mechanism for most cellular mRNAs. The 43S scans along the 5′UTR of the mRNA.

Upon reaching the initiation codon, the 60S subunit joins and translation factors are released to form the 80S ribosome (Jackson, Hellen et al. 2010). The ribosome then proceeds to the elongation stage of translation until it reaches a stop codon on the mRNA.

Given the complexity of initiation, this step is regulated by many proteins and signaling cascades (Figure I.3).

14 15

Figure I.3: Major steps of cap-dependent translation initiation

Depicted is a simplified model of translation initiation and its regulation by two major signaling pathways, the integrated stress response (ISR) and mTOR. A) The ternary complex is composed of eIF2, GTP, and the methionyl tRNA for initiation. The ternary complex and 40S ribosomal subunit joins to form the 43S preinitiation complex (PIC). B) mRNA activation is mediated by the cap-binding complex eukaryotic initiation factor (eIF) 4F. eIF4F (composed of the scaffold eIF4G, the cap-binding protein eIF4E, and the RNA helicase eIF4A) binds the 5′ m 7G cap and poly(A) binding protein (PABP) to facilitate mRNA circularization. C) The PIC is recruited to the 5′ end of the activated mRNA by eIF4F, and after attachment to the mRNA, the PIC scans along the 5′UTR until it reaches the initiation codon. During the ISR, phosphorylation of eIF2 by kinases reduces formation of the PIC to globally decrease translation initiation. mTOR signaling converges on phosphorylation of several eIFs to facilitate mRNA activation and PIC scanning on the 5′UTR to globally promote translation initiation. Some eukaryotic initiation factors are excluded from the PIC for simplicity.

15 16

mTORC1 is a major regulator of translation initiation and required for multiple forms of synaptic plasticity. Neurotrophic factors and glutamate can activate the PI3K-

AKT and MEK/ERK signaling cascade to converge on TSC-Rheb-mTORC1. Activation of multi-unit kinase mTORC1 leads to phosphorylation of multiple downstream substrates to ultimately promote initiation (Figure I.4A). Two major mTORC1 substrates are the translation repressor eukaryotic translation initiation factor 4E-binding protein

(4E-BP) and ribosomal protein S6 kinase beta-1 (S6K1). 4E-BP normally binds the cap- binding protein eIF4E to impair cap-dependent translation, and phospho-4E-BP has decreased affinity for eIF4E. mTORC1 activates S6K1, leading to subsequent phosphorylation of ribosomal protein S6 (rpS6), eukaryotic elongation factor 2 kinase

(eEF2K) and eukaryotic initiation factor 4B (eIF4B). Phospho-rpS6 is a widely used readout for mTORC1 activity, but the impact of phospho-rpS6 on translation is unclear.

Phosphorylation of eEF2K and eIF4B broadly supports increased protein synthesis by promoting elongation and translation of mRNAs with long, structured 5′UTRs (Gandin,

Masvidal et al. 2016).

The role of mTOR signaling in synaptic plasticity and learning has been demonstrated through chemical inhibition of the pathway with rapamycin and mouse models of upstream regulators and downstream effectors. Rapamycin was initially discovered as an anti-fungal isolated from bacteria in soil from the island of Rapa Nui

(Vézina, Kudelski et al. 1975). Rapamycin inhibits cell proliferation by blocking protein synthesis (Martel, Klicius et al. 1977, Singh, Sun et al. 1979). Rapamycin binds to FK506 binding protein 12 (FKBP12) to predominantly inhibit mTORC1, but rapamycin can also

16 17

impair mTORC2 when applied in higher concentrations for longer duration (Sarbassov,

Ali et al. 2006). Rapamycin impairs several forms of protein-synthesis dependent synaptic plasticity in the hippocampus, such as BDNF-induced LTP and mGluR-LTD

(Tang, Reis et al. 2002, Hou and Klann 2004). Furthermore, injection of rapamycin into the hippocampus impairs protein-synthesis dependent long-term spatial memory formation in mice as measured by the Morris water maze, an assay where mice remember the location of a hidden platform based on visual cues (Dash, Orsi et al. 2006).

Knockout mouse models of the major mTORC1 substrates 4E-BP2 and S6K1/2 further support a role of mTORC1-dependent processes in synaptic plasticity and learning. mTOR and 4E-BP1/2 are present in the stratum radiatum (a region enriched for dendrites) of the hippocampus and synapses in cultured neurons (Tang, Reis et al. 2002).

Eif4ebp2-KO mice also have altered synaptic plasticity (deficient L-LTP, enhanced mGluR-LTD) and deficits in protein-synthesis dependent forms of learning (long-term spatial learning and fear conditioning) (Banko, Poulin et al. 2005, Banko, Hou et al.

2006). Mice lacking S6K1 or S6K2 have altered contextual fear memory and spatial memory (Antion, Hou et al. 2008). In addition, mice lacking TSC, the upstream inhibitor of mTOR, have learning deficits and evidence of increased S6K signaling consistent with mTOR hyperactivation (Ehninger, Han et al. 2008).

mTORC1 can also increase the translation capacity of cells through increased protein synthesis of components of the translation machinery. A group of mRNAs sensitive to mTORC1 regulation is the 5′ terminal oligopyrimidine tract (TOP) mRNAs, which mostly encode ribosomal proteins and translation factors. Translation of TOP

17 18

mRNAs is regulated through phosphorylation of 4E-BP and the RNA-binding protein

LARP1 (Thoreen, Chantranupong et al. 2012, Hong, Freeberg et al. 2017, Philippe,

Vasseur et al. 2017). Dendritic translation of TOP mRNAs has been demonstrated following some forms of synaptic plasticity on a single gene level, such as eEF1A following mGluR5 signaling (Huang, Chotiner et al. 2005). However, it is unclear whether as a group the TOP mRNAs are translated during synaptic plasticity in an mTORC1-dependent manner.

The experiments with rapamycin and knockout mouse models of mTORC1 effectors support a role of mTORC1 signaling, 4E-BP, and S6K in translation of proteins required for synaptic plasticity and higher cognitive function. However, they do not exclude mTORC1 effects outside of translation (such as autophagy and lipogenesis) or mTORC1 independent effects of the 4E-BPs and S6K. Recent work has highlighted the importance of mTORC2, a more poorly characterized complex that cross talks with mTORC1 signaling to regulate the actin cytoskeleton. Somewhat contradictory to the experiments described earlier, Zhu et al argue that mTORC2, rather than mTORC1, is the major effector of mGluR-LTD (Zhu, Chen et al. 2018).

Another major regulator of translation initiation required for synaptic plasticity is the integrated stress response (ISR) (Figure I.4B). Whereas mTOR signaling promotes translation to support cell proliferation, the ISR allows cells to adapt to a stress by inhibiting general protein synthesis while promoting translation of specific mRNAs

(Hinnebusch 2005). The ISR converges on phosphorylation of the alpha subunit of the eIF2 ternary complex (eIF2α), which can be phosphorylated by one of four kinases

18 19

(GCN2, PERK, PKR, and HRI) in response to a variety of stressors. eIF2α is phosphorylated on Ser51, which impairs the formation of the ternary complex (eIF2,

Met-tRNA, GTP) and subsequently reduces initiation for most mRNAs. Instead, the ISR promotes translation of mRNAs containing up-stream open reading frames (uORFs).

Recent work has highlighted the role of the ISR in synaptic plasticity and higher cognitive function (Costa-Mattioli, Gobert et al. 2007, Di Prisco, Huang et al. 2014). mGluR5 signaling induces eIF2α phosphorylation and translation of a uORF-containing mRNA, Ophn1 (Di Prisco, Huang et al. 2014). eIF2α phosphorylation is required for mGluR-LTD, because chemical inhibition of eIF2α phosphorylation and a phosphorylation site mutation impairs mGluR-LTD.

19 20

Figure I.4: Overview of translation signaling pathways important for mGluR-LTD

A) mGluR5 signaling activates mTORC1, which phosphorylates its major substrate S6K and 4E-BP. S6K phosphorylates multiple substrates, such as RPS6 and eIF4B, which promotes recruitment of the 43S preinitiation complex to mRNAs. Phospho- 4E-BP has lower affinity for eIF4E (the cap-binding protein) and promotes cap- dependent translation. mRNAs that contain a terminal oligopyrimidine (TOP) tract immediately after the 5′ cap are particularly sensitive to mTORC1-dependent signaling. Collectively, mTORC1 signaling promotes global protein synthesis, whereas the chemical inhibitor rapamycin inhibits mTORC1-dependent translation. B) mGluR5 signaling activates the integrated stress response, which converges on phosphorylation of eIF2α (part of the ternary complex required for initiation), by kinases. Phospho-eIF2α decreases global protein synthesis, but promotes translation of mRNAs containing upstream open reading frames (uORFs). C) mGluR5 signaling activates eEF2K, the sole kinase that phosphorylates eukaryotic elongation factor 2 (eEF2). Phospho-eEF2 has decreased binding to ribosome and thus cannot promote ribosome translocation, leading to inhibition of global translation. Through a poorly characterized mechanism, phospho-eEF2 can also promote protein synthesis of poorly translated mRNAs required for mGluR-LTD. It is unknown how these multiple translation pathways interact to regulate synthesis of proteins required for mGluR-LTD. Positive regulators (green) and negative regulators (red) of translation indicated. Yellow circle with P represents phosphorylation events.

20 21

Regulation of translation elongation in synaptic plasticity Translational regulation has historically focused on initiation, but emerging studies reveals that translation elongation is also important for synaptic plasticity and appropriate brain function. Compared to translation initiation, translation elongation is a much simpler process. After initiation, eukaryotic elongation factor 1 (eEF1) bound to

GTP brings the tRNA into the empty A site of the ribosome. eEF1-GTP is hydrolyzed to eEF1-GDP, and the growing nascent peptide is transferred to the A-site tRNA. Next, eukaryotic elongation factor 2 (eEF2) bound to GTP enters the A site to catalyze the transfer of the A-site peptidyl-tRNA to the P site. eEF2-GDP is released, and a round of elongation can repeat.

Regulation of translation elongation occurs through phosphorylation of eEF2 on

Thr56 within the GTP-binding domain, leading to global reduction of protein synthesis

(Ryazanov, Shestakova et al. 1988, Price, Redpath et al. 1991). This phosphorylation event likely causes structural rearrangement of eEF2 to reduce binding to GTP and/or eEF2-GTP binding to ribosomes, leading to a roughly 10-100 times decreased affinity of eEF2 for the ribosomes without affecting GTP hydrolysis (Ryazanov, Shestakova et al.

1988, Carlberg, Nilsson et al. 1990, Nilsson and Nygård 1991, Dumont-Miscopein,

Lavergne et al. 1994). Eukaryotic elongation factor 2 kinase (eEF2K), a calcium/calmodulin-activated kinase (CaMKIII), phosphorylates eEF2, its sole substrate. eEF2K signaling is activated following different types of neuronal stimulation and is required for synaptic plasticity and learning and memory (Sutton et al 2007). mGluR-

LTD signaling leads to eEF2 phosphorylation, and mice lacking eEF2K have impaired

21 22

mGluR-LTD (Figure I.4C) (Park, Park et al. 2008). mGluR5/eEF2K signaling is required for translation of specific mRNAs, such as Arc and Bdnf, to support mGluR-LTD (Park,

Park et al. 2008, Verpelli, Piccoli et al. 2010). The exact mechanism of how phospho- eEF2 can promote translation of specific mRNAs while inhibiting general protein synthesis is not well understood. One model proposes that slowing ribosome translocation can free up initiation factors to promote synthesis of inefficiently translated mRNAs (Walden and Thach 1986, Scheetz, Nairn et al. 2000). Further support comes from experiments with low dose cycloheximide, which can paradoxically increase translation of mRNAs (e.g. CaMKII, Arc) required for synaptic plasticity (Scheetz, Nairn et al. 2000, Park, Park et al. 2008).

Recent work has highlighted the importance of translation elongation in the brain for neurodevelopmental and neurodegenerative disorders. Mutations in eukaryotic elongation factor 1A2 (eEF1A2) have been identified in patients with atypical Rett

Syndome and patients with intellectual disability, autism, and seizures (Nakajima,

Okamoto et al. 2015, Inui, Kobayashi et al. 2016). Regulation of mRNAs at elongation may allow for more rapid protein synthesis than initiation. A recent publication found that mRNAs with stalled and slow moving ribosomes in the hippocampus are highly enriched for neuronal function, such as axonal development and synaptic organization

(Shah, Molinaro et al. 2020). Several elegant studies indicate that synaptic activity reactivates stalled ribosomes/granule complexes in dendrites. b-actin mRNA is unmasked in dendrites following a chemical LTP stimulation, suggesting that mRNA-ribosome complexes are stored in a repressed state until neuronal activation (Buxbaum and Singer

22 23

2014). Furthermore, Graber et al surprisingly found that harringtonine, a chemical that prevents the first round of elongation by blocking peptide bond formation without affecting downstream ribosomes, does not impair mGluR-LTD (Huang 1975, Fresno,

Jiménez et al. 1977). This argues that at least some mRNAs required for mGluR-LTD are associated with stalled ribosomes downstream of the first round of elongation (Graber,

Hebert-Seropian et al. 2013).

Convergence of mGluR5 and mTOR signaling on ASD Given the heterogeneity and numerous genes associated with ASD, current studies are focused on identifying and understanding common pathways dysregulated across multiple forms of ASD. One such example is mTOR and mGluR5 signaling.

Mutations in multiple upstream regulators of mTOR signaling and downstream effectors are high-risk autism mutations in humans (Figure I.5). TSC and PTEN are leading genetic disorders linked to syndromic ASD, and mutations in other components of mTOR signaling (MTOR, RHEB, NF1, S6K) have been identified in idiopathic ASD.

Furthermore, altered mTOR signaling is observed in mouse models of other syndromic causes of ASD, including FXS, Rett Syndrome, and Phelan-McDermid syndrome.

Multiple mouse models of ASD have altered mGluR-LTD. Huber et al first described that DHPG-induced mGluR-LTD is increased in a mouse model of FXS

(Huber, Gallagher et al. 2002). Furthermore, the LTD in Fragile X mice was protein synthesis independent, because the translation inhibitor cycloheximide had no effect on

LTD. Fragile X mice have increased protein synthesis in the brain, which correlates with increased PI3K/AKT/mTOR signaling (Sharma, Hoeffer et al. 2010). Multiple studies

23 24

suggest that elevated mTOR signaling contributes to dysregulated translation and FXS phenotypes. Protein synthesis in the brain, synaptic plasticity, and behavior are rescued in the FXS mouse model by chemical inhibition or genetic removal of S6K1 (encoded by

Rps6kb1), a major mTORC1 substrate (Bhattacharya, Kaphzan et al. 2012). Although it is uncertain if the increased mTOR signaling is causative or a response in FXS mice, the

Fmr1/Rps6kb1-KO mice suggests restoring mTORC1/S6K signaling rescues phenotypes of Rps6kb1-KO mice and male mice lacking Fmr1 (Fmr1-/y). Numerous FXS rescue mouse models have found a correlation between restoring protein homeostasis and correcting mGluR-LTD synaptic plasticity and autism-like behavior (Dölen, Osterweil et al. 2007, Auerbach, Osterweil et al. 2011, Bhattacharya, Kaphzan et al. 2012, Udagawa,

Farny et al. 2013, Gross, Chang et al. 2015). Collectively, these experiments led to the

“mGluR5 theory” of Fragile X Syndrome, where excessive mGluR5 signaling may drive the excessive protein synthesis underlying the altered synaptic plasticity (Bear, Huber et al. 2004). The mGluR theory of Fragile X Syndrome led to the development of mGluR5 inhibitors for use in clinical trials. In addition, modulators of mGluR5 have shown promise in reducing seizures, ameliorating ASD-like behavior, and restoring protein homeostasis in a TSC mouse model of autism (Kelly, Schaeffer et al. 2018).

Given that TSC is an inhibitor of mTOR signaling, one predicts that mice haploinsufficient in TSC would have increased mGluR-LTD similar to the FXS mice.

Surprisingly, mice and rats with Tsc1 or Tsc2 deficiency consistently have decreased mGluR-LTD (von der Brelie, Waltereit et al. 2006, Auerbach, Osterweil et al. 2011). In a

Tsc2 heterozygous mouse model (Tsc2+/-), this correlates with decreased protein

24 25

synthesis as measured by 35S-methionine labeling in the hippocampus (Auerbach,

Osterweil et al. 2011). Rapamycin can restore mGluR-LTD to wild-type levels in Tsc2+/- mice, arguing that excessive mTOR signaling underlies the synaptic plasticity phenotype.

The mechanism of how Tsc2+/- mice have decreased mGluR-LTD and protein synthesis is unknown. Regardless, a combined Tsc2+/- and Fmr1-/y mouse also restores protein synthesis in the brain and mGluR-LTD to wild-type levels (Auerbach, Osterweil et al.

2011). These results support a model where a common set of mRNAs or pathways are altered in TSC and FXS. My project sought to identify the mRNAs whose translation is responsive to mGluR5 signaling and dysregulated in a mouse model of TSC by using ribosome profiling, an unbiased, high-resolution method to measure translation.

25 26

Figure I.5: mGluR5 and mTORC1 signaling to translation

DHPG binding to mGluR5 leads to activation of MEK/ERK and PI3K/AKT signaling. ERK and AKT signaling converges to inhibit TSC (composed of TSC1/2), allowing for RHEB to activate mTORC1 (composed of core components mTOR, RAPTOR, mLST8, and other accessory proteins excluded for simplicity). mTORC1 is a serine/threonine kinase that inactivates the translational repressor 4E-BP and activates S6K1 to broadly promote cap-dependent protein synthesis. FMRP is degraded following mGluR5 signaling, relieving translational repression on its targets. The stars indicate mutations in genes associated with autism in humans (gene.sfari.org). Positive regulators (green) and negative regulators (red) of translation are indicated.

26 27

CHAPTER II RIBOSOME PROFILING IN MOUSE HIPPOCAMPUS: PLASTICITY-INDUCED REGULATION AND BIDIRECTIONAL CONTROL BY TSC2 AND FMRP

Annie Hien1*, Gemma Molinaro2*, Botao Liu1, Kimberly M. Huber2, Joel D. Richter2

1Program in Molecular Medicine

University of Massachusetts Medical School

Worcester, MA 01605

2Department of Neuroscience

University of Texas Southwestern Medical Center

Dallas, TX 75390

*equal contributors

27 28

Preface

K.M.H. and J.D.R. conceived the study. G.M. collected and prepared the hippocampal slices for sequencing, performed western blotting, and performed e- physiology experiments in Figure II.4B, Figure II.7C, Figure II.12, and Figure II.14. A.H. prepared sequencing libraries for the Tsc2+/- mice and their wild-type littermates. B.L. prepared 4-5 sequencing libraries from wild-type samples before and after DHPG treatment that were used in Figure II.8 and downstream analyses. A.H. performed all sequencing data analysis and prepared all the figures, except for Figure II.12 and Figure

II.14, which were compiled and figure legend written by K.M.H. and G.M. A.H., J.D.R., and K.M.H. interpreted the results and wrote the manuscript with input from all authors.

This work was submitted for publication in Molecular Autism and is currently under revision.

Annie Hien, Gemma Molinaro, Botao Liu, Kimberly Huber, and Joel Richter.

Ribosome Profiling in Mouse Hippocampus: Plasticity-induced Regulation and

Bidirectional control by TSC2 and FMRP. (2020)

Introduction

Tuberous sclerosis (TSC) is an autosomal dominant disorder characterized by benign tumor growth in multiple organs and neuropsychiatric symptoms. Individuals with

TSC have an increased incidence of seizures (~90%), intellectual disability (~50%), and autism (~50%) (Lipton and Sahin 2014). Disrupted neuronal circuitry likely underlies many neuro-pathologies in TSC because individuals with an anatomically normal brain 28 29

can still present substantial developmental delay, intellectual disability, and autism. More direct evidence for impaired neuronal connectivity is derived from mouse models of the disorder, which display hyperexcitability, aberrant synaptic plasticity, and altered dendritic spine morphology (Tavazoie, Alvarez et al. 2005, Auerbach, Osterweil et al.

2011). TSC is caused by loss-of-function mutations in TSC1 or TSC2; mutations in TSC2 are more common and are responsible for the most severe symptoms. The TSC1/2 proteins heterodimerize to form the tuberous sclerosis complex (TSC), a GTPase- activating protein that inhibits Ras homology enriched in brain (Rheb) (Garami,

Zwartkruis et al. 2003, Zhang, Gao et al. 2003). Rheb activates the mechanistic target of rapamycin (mTOR), a kinase that forms two biochemically distinct complexes: mTORC1 and mTORC2 (Saxton and Sabatini 2017). mTORC1 regulates mRNA translation primarily via phosphorylation of 4E-BP and the kinase SK1 to phosphorylate rpS6

(Brunn, Hudson et al. 1997, Gingras, Kennedy et al. 1998, Shaw and Cantley 2006). The less-characterized mTORC2 plays a role in actin cytoskeletal reorganization (Saxton and

Sabatini 2017). Complete loss of TSC1 or TSC2 leads to excessive mTOR activity, and mutations in other components of the mTOR pathway are also linked to autism in humans such as PTEN, RHEB, and MTOR (O'Roak, Vives et al. 2012, Iossifov, O'Roak et al.

2014). Thus, the mTOR pathway forms a highly connected signaling network linked to autism. Although pharmacological options are available for the treatment of seizures in individuals with TSC, there currently is no effective pharmacological treatment for the neuropsychiatric symptoms of intellectual disability and autism.

29 30

The mTOR pathway integrates signaling from various inputs to regulate synaptic plasticity and higher cognitive function. One example is metabotropic glutamate receptor-mediated long-term depression (mGluR-LTD), a form of synaptic plasticity that requires de novo protein synthesis (Huber, Kayser et al. 2000). The mGluR1/5 agonist dihydroxyphenylglycine (DHPG) stimulates group 1 metabotropic glutamate receptors to activate the PI3K/AKT/mTOR pathway, leading to translation of largely unidentified mRNAs required for mGluR-LTD (Hou and Klann 2004, Ronesi, Collins et al. 2012).

Multiple mouse models of autism, such as TSC and Fragile X Syndrome, have altered hippocampal mGluR-dependent synaptic plasticity (Huber, Gallagher et al. 2002,

Auerbach, Osterweil et al. 2011, Bateup, Takasaki et al. 2011, Auerbach, Osterweil et al.

, Chévere-Torres, Kaphzan et al. 2012). Although experiments with rapamycin and inhibitors of protein synthesis suggest that excessive mTORC1-dependent translation drives the altered synaptic plasticity phenotype in TSC mice, other findings suggest a more complex picture. For example, Tsc2+/- mice have reduced hippocampal protein synthesis rates instead of the predicted excessive protein synthesis from mTOR hyperactivation (Auerbach, Osterweil et al. 2011). In addition, a recent study challenged the role of mTORC1 in hippocampal mGluR-LTD, arguing that mTORC2 is the major regulator of this form of synaptic plasticity (Park, Park et al. 2008, Zhu, Chen et al. 2018) but see (Hou and Klann 2004, Sharma, Hoeffer et al. 2010). MGluR1/5 signaling can also regulate translation factors to suppress global protein synthesis, such as phosphorylation of eukaryotic initiation factor 2 alpha and eukaryotic elongation factor 2 (eEF2), which are both necessary for mGluR-LTD (Park, Park et al. 2008, Di Prisco, Huang et al. 2014).

30 31

Although mGluR1/5 stimulation is known to promote translation of specific mRNAs

(Waung, Pfeiffer et al. 2008), how translation is regulated in a transcriptome-wide manner by mGluR1/5 in either the normal or TSC brain is unknown.

Using a germline Tsc2+/- mouse model of TSC autism, we performed ribosome profiling, an unbiased whole transcriptome method that determines the number and positions of ribosomes on all mRNAs, and thus serves as a proxy for protein synthesis

(Ingolia, Ghaemmaghami et al. 2009). Together with RNA-seq, ribosome profiling in wild-type and Tsc2+/- hippocampal slices under basal conditions and after induction of mGluR-LTD allows us to define posttranscriptional changes in gene expression in response to synapse activation (Huber, Kayser et al. 2000). We chose the Tsc2+/- mouse model because we strove to capture translational dysregulation that is reflective of TSC autism arising from TSC2 haploinsufficiency in humans. This mouse model displays excessive mTOR signaling, a well-characterized deficient mGluR synaptic plasticity, and alterations in social interactions and hippocampal-dependent learning tasks that are reversed by rapamycin treatment (Hernandez, Way et al. 2007, Ehninger, Han et al. 2008,

Auerbach, Osterweil et al. 2011, Sato, Kasai et al. 2019). Using ribosome profiling, we identified changes in the Tsc2+/- hippocampus as well as following mGluR-LTD signaling. We also observed that mGluR1/5 stimulation induces a rapid increase in mRNA abundance, and targets of Fragile X Mental Retardation Protein (FMRP), an

RNA-binding protein whose loss causes Fragile X Syndrome, are enriched in this group. mGluR stimulation also promotes phosphorylation of eukaryotic elongation factor

(eEF2), which is deficient in Tsc2+/- mice. These animals have a reduced mGluR-LTD

31 32

and chemically mimicking phospho-eEF2 enhances this form of synaptic plasticity. Our results suggest a novel pathway of translational dysregulation in ribosome translocation

(polypeptide elongation) that may contribute to altered translation and synaptic plasticity in TSC mice.

Results

Translational Profiling of the TSC hippocampus The GTPase activating protein tuberous sclerosis complex (TSC) is an upstream but indirect repressor of mTOR, a kinase which forms the multi-subunit complex mTORC1 to stimulate translation in response to a number of signaling inputs including activated group I metabotropic glutamate receptors (e.g., mGluR5). mGluR5 activation leads to long term depression (LTD), a protein synthesis-dependent form of synaptic plasticity frequently examined at hippocampal Schafer collateral-CA1 synapses (Huber,

Kayser et al. 2000). To model TSC, a multi-organ disorder characterized by autism, epilepsy, and intellectual disability, we generated mice lacking one copy of the Tsc2 gene, which will henceforth be referred to as the Tsc2+/- mouse.

32 33

To detect subtle changes in translation, we turned to ribosome profiling, a high- resolution and unbiased whole transcriptome method for determining the number and positions of ribosomes on mRNA. When combined with RNA-seq, the ratio of ribosome occupancy over input mRNA yields translational efficiency (TE) (Ingolia,

Ghaemmaghami et al. 2009). Changes in TE are often used as a proxy for protein synthesis, where an increased TE suggests increased translation rates and protein production. However, changes in TE may not always reflect protein output when ribosome elongation or mRNA levels are altered. Hippocampal slices, some of which were treated with dihydroxyphenylglycine (DHPG) for 5 min to induce LTD, were lysed in the presence of cycloheximide to prevent ribosome run-off, digested with RNases A and T1, and the 80S monomers containing ~30nt ribosome protected fragments (RPFs) were collected by ultracentrifugation (Figure II.1). The isolated RPFs were appended with primers for library construction and sequencing (>10 million uniquely mapped

Figure II.1: Schematic of experimental design

RNA-Seq and ribosome profiling were performed from CA1-enriched hippocampal slices of Tsc2+/- and wild-type littermate controls. Slices were treated with DHPG (100µM, 5min) or artificial cerebrospinal fluid (vehicle) and then processed for RNA- Seq and ribosome profiling. Slices from 6 mice were pooled per genotype and condition, and 3 biological replicates were performed. RPF= ribosome protected fragment.

33 34

reads; libraries were strongly correlated with each other; Pearson’s R>0.95).

Figure II.2A shows relative expression levels of the differentially expressed (DE) genes with altered levels of RNA and RPFs from the hippocampal slices of the Tsc2+/- mice and their littermate controls under basal (unstimulated) conditions that pass a significance cutoff of FDR < 0.1. Tsc2 was the top differentially expressed mRNA at both the RNA and RPF level, confirming the validity of our approach (Figure II.2B). The

DE mRNAs with altered levels of RNA and RPF in Tsc2+/- had little overlap with each other (n = 5). The ratio of RPF to RNA yields translational efficiency (TE), but because changes in TE do not distinguish between RPF and RNA effects and because few RNAs passed statistical cutoff for changes in TE (log2FC), we considered RNA and RPF changes separately.

We observed enrichment for terms related to the GABA-receptor complex, mitochondrial membrane, and myelin sheath for RPFs that were Up regulated in the

Tsc2+/- brain (Figure II.2C). No GO terms were enriched in the Down regulated RPF group, likely due to the small number of genes. An altered excitatory/inhibitory balance is frequently observed in ASD, and the increased translation of GABA-A related proteins might be a compensatory response to the increased hyperexcitability observed in Tsc2+/- mice (Bateup, Johnson et al. 2013, Potter, Basu et al. 2013, Nelson and Valakh 2015,

Antoine, Langberg et al. 2019, Basu, Riordan et al. 2019). Altered translation of mitochondrial membrane proteins is consistent with the role of mTOR-dependent translational control of mitochondrial-related mRNAs and reports of mitochondrial dysfunction in Tsc2-deficient cultured neurons (Morita, Gravel et al. 2013, Ebrahimi-

34 35

Fakhari, Saffari et al. 2016). The abnormalities in myelination in TSC patients and function of TSC/mTOR in myelination and function of oligodendrocytes (Tsai and Sahin

2011, Peters, Struyven et al. 2019) may be due to translational control of myelin sheath proteins.

35 36

Figure II.2: Ribosome profiling of Tsc2+/- hippocampal slices reveals RNA and RPF dysregulation

A) Heatmaps show differentially expressed genes with altered levels of RNAs (left) and RPFs (right) in Tsc2+/- compared to WT under basal conditions. Relative gene expression is shown for each mRNA (row), and each column represents a biological replicate (n=3) for RNA (DE=109, FDR < 0.1) and RPFs (DE=99, FDR < 0.1). B) Normalized count plots of RNA (top) and RPF (bottom) levels for Tsc2 are shown (Wald-test for genotype comparison with BH correction. RNA: log2FC -0.52, padj < 7.4e-32; RPF log2FC -0.70, padj < 2.4e-22). C) GO term enrichment of mRNAs with upregulated RPFs in Tsc2+/- compared to WT (padj < 0.05).

36 37

Dysregulation of 5′ terminal oligopyrimidine (TOP) mRNAs in Tsc2+/- The major downstream substrate of TSC is mTOR, which controls specific and global translation (Figure II.3, top). mRNAs that contain a 5′ terminal oligopyrimidine motif (TOP), which generally encode components of the translational machinery, and mRNAs with long structured 5′UTRs are especially sensitive to mTOR regulation

(Svitkin, Pause et al. 2001, Holz, Ballif et al. 2005, Thoreen, Chantranupong et al. 2012,

Gandin, Masvidal et al. 2016). TSC/mTOR also regulates general translation through phosphorylation and dephosphorylation events such as those that occur on 4E-BP, eIF4B, rpS6, and eEF2 (Redpath, Foulstone et al. 1996, Browne and Proud 2004, Raught,

Peiretti et al. 2004, Holz, Ballif et al. 2005). Elevated mTOR signaling contributes to various behavioral and synaptic plasticity phenotypes in the brains of TSC mice as evidenced by their rescue by treatment with the mTOR inhibitor rapamycin (Ehninger,

Han et al. 2008, Auerbach, Osterweil et al. 2011). We found that the TOP mRNAs had mildly increased RPFs, but as a group, they had increased TE that is primarily driven by decreased RNA levels (Figure II.3, bottom). mTOR can also promote initiation of non-

TOP mRNAs that have long, structured 5′UTRs via eIF4E and eIF4A (Gandin, Masvidal et al. 2016). However, we find no difference in the 5′UTR lengths or predicted secondary structures in the up- and down-regulated RPFs (Figure II.4A). To assess whether signaling events usually associated with mTOR activation occur in Tsc2+/- hippocampus, we examined phospho-mTOR and phospho-rpS6, which are indirect downstream substrates of TSC. Changes in the basal levels of phospho-mTOR are undetectable in

37 38

TSC hippocampus, and there is only a trend towards increases in phospho-rpS6 (Figure

II.4B), similar to previous reports (Potter, Basu et al. 2013).

Figure II.3: Dysregulation of TOP mRNAs in Tsc2+/-

Top) Schematic depicting mTOR control of translation. mTOR controls mRNA- specific translation via 5′ terminal oligopyrimidine (TOP) motifs or long, structured 5′UTRs. mTOR controls global initiation and elongation via phosphorylation/dephosphorylation of 4E-BP, rpS6, and eEF2. Bottom) Density plots depict distribution of mRNA, RPF, and translation efficiency (TE) expression changes (Tsc2+/-/WT log2FC) for TOP mRNAs (n=68) compared to non-TOP mRNAs. Wilcoxon rank-sum test with correction by Bonferroni method.

38 39

Figure II.4: Lack of unique 5′UTR features in dysregulated RPFs from Tsc2+/-

A) Density plots depict the distribution of the length (left) and minimal free energy (MFE) (right) of the 5′UTR of Down and Up RPFs in Tsc2+/- compared to all mRNAs. ns = non-significant, Wilcoxon rank-sum test with correction by Bonferroni method. B) The levels of phosphorylated and total rpS6 (left) and mTOR (right) as determined by western blots of fresh, whole hippocampus from wild-type and Tsc2+/- littermate (n= 5 animals, ns = non-significant, unpaired t-test). Levels of total rpS6 are not significantly altered between Tsc2+/- and wild-type hippocampus (88+/- 13% of WT; n = 5 mice/genotype; unpaired t-test).

39 40

Because we did not identify specific mRNA features in the mRNAs and RPFs dysregulated in Tsc2+/- mice, we wondered if the mRNA-specific dysregulation may arise by decreased global protein synthesis as previously demonstrated in these mice

(Auerbach, Osterweil et al. 2011). Mills and Green proposed the “ribosome concentration” hypothesis to explain how a modest reduction in global protein synthesis rates from decreased ribosome levels might differentially affect translation of mRNAs with low and high initiation rates (Mills and Green 2017). Based on this model, translation of most mRNAs would be unaffected, but mRNAs with high initiation rates may have increased protein output because of decreased inhibitory ribosome crowding.

Conversely, mRNAs with low initiation rates, such as those with long, structured

5′UTRs, uORFs, internal ribosome entry sites, or poor Kozak sequences, would have decreased protein output because these mRNAs are poor substrates for translation initiation. Because Tsc2+/- mice have a roughly 25% decrease in global protein synthesis through an unknown mechanism, we used the “ribosome concentration” model as an intellectual framework for interpreting our results (Auerbach, Osterweil et al. 2011). We approximated initiation by calculating TE from wild-type basal slices, which will be referred to as hippocampal TE. The Up RNAs in Tsc2+/- had a low hippocampal TE compared to all (Figure II.5, left). We find that mRNAs with Down RPFs in Tsc2+/- had a low hippocampal TE and the Up RPFs had a high hippocampal TE compared to all

(Figure II.5, middle). These results suggest that in Tsc2+/- mice, poorly translated mRNAs are reduced even further while highly translated mRNAs become more robustly translated, consistent with the “ribosome concentration” hypothesis (Figure II.5, right).

40 41

Finally, Figure II.6 shows transcript features that correlate with hippocampal TE.

Consistent with another study, hippocampal TE is correlated with GC content of the

CDS, 3′UTR, and 5′UTR and also CDS length (Chan, Mugler et al. 2018).

Figure II.5: Correlation of hippocampal TE with degree of dysregulation in Tsc2+/-

Density plots depict distribution of the hippocampal TE for the up and down RNA (left) and RPF (middle) groups in Tsc2+/-/WT compared to all genes (one-tailed Kolmogorov-Smirnov test, pvalues indicated in plots). The hippocampal TE is calculated from the average of wild-type basal slices as the ratio of RPF to RNA. Right: Schematic depiction of how loss of Tsc2 could affect translation of lowly and highly translated mRNAs.

41 42

Figure II.6: Transcript features of hippocampal TE

Pearson correlation coefficient was computed for pairs of transcript features including translation efficiency (TE), coding length and GC content (CDS), 5′ untranslated region length and GC content (5′UTR), 3′ untranslated region length and GC content (3′UTR), and RNA abundance (TPM, transcripts per million). Red indicates a positive correlation, and blue a negative correlation. Gray indicates a correlation of identical parameters. p < 0.05 for all correlation coefficients.

42 43

FMRP binding targets are altered in opposite directions in TSC and FXS mice Loss of the RNA-binding protein Fragile X Mental Retardation Protein (FMRP) leads to Fragile X Syndrome (FXS). FMRP and TSC2 bidirectionally regulate mGluR-

LTD and behavior, suggesting they might regulate similar mRNAs. In support of this, a combined TSC/FXS mouse rescues deficits in synaptic plasticity and learning to wild- type levels compared to the single mouse mutants (Auerbach, Osterweil et al. 2011).

Given the rescue paradigm of the FXS and TSC mouse, we determined if FMRP targets are altered in Tsc2+/- mice. We used a list of FMRP targets identified by CLIP (crosslink and immunoprecipitation) in CA1 pyramidal neurons, where FMRP targets were stratified into 3 groups based on binding strength: stringent (n=327), high (n=938), and low (n=1330) binding targets (Sawicka, Hale et al. 2019). FMRP targets are enriched (p <

5.3e-8) in the mRNAs altered in TSC, and as a group, they have decreased TE that is driven by increased RNA levels that correlate with stringency of FMRP binding (Figure

II.7A). FMRP targets have no change in RPFs in TSC (Figure II.7A, middle).

Interestingly, the FMRP targets are decreased in the FXS mouse (Fmr1-/y) at the RNA and RPF level, which is also correlated with stringency of FMRP binding (Figure

II.7B)(Ceolin, Bouquier et al. 2017, Sawicka, Hale et al. 2019, Shu, Donnard et al. 2019,

Shah, Molinaro et al. 2020). Because FMRP binding targets were dysregulated in Tsc2+/-, we determined if FMRP protein levels were altered in Tsc2+/-. FMRP levels were unaltered in Tsc2+/- (Figure II.7C), suggesting that altered FMRP levels are not responsible for the increased RNA levels of FMRP targets in Tsc2+/- mice. This bidirectional change in FMRP target mRNAs could be linked to the TSC/FXS mouse

43 44

rescue paradigm and reveals a complex interaction of TSC2 and FMRP in steady-state regulation of mRNAs in the hippocampus.

44 45

Figure II.7: FMRP binding targets are regulated by TSC2 and FMRP in opposing directions

Empirical Cumulative Distribution Function (ECDF) of RNA (left), RPF (middle), and TE (right) log2FC expression changes of FMRP binding targets compared to non- target mRNAs in A) Tsc2+/-/WT and B) Fmr1-/y/WT mice. FMRP binding targets are stratified into 3 groups based on binding stringency. Wilcoxon rank-sum test with Bonferroni correction. Data of hippocampal slices of Fmr1-/y mice from Shah et al (Shah, Molinaro et al. 2020) and FMRP binding targets from Sawicka et al (Sawicka, Hale et al. 2019). C) The levels of FMRP as determined by western blots of fresh, whole hippocampus from wild-type and Tsc2+/- littermates. n= 5 animals, ns = non-significant, unpaired t-test.

45 46

mGluR5 signaling induces a rapid RNA and translational response mGluR-LTD is regulated by TSC, so to determine the identity of mGluR- stimulated translation of specific RNAs during LTD, we performed ribosome profiling and RNA-seq on hippocampal slices following 5 minutes of DHPG treatment (Huber,

Kayser et al. 2000, Park, Park et al. 2008, Waung, Pfeiffer et al. 2008). Figure II.8A shows that RNA (DE=241, FDR < 0.01) and RPF (DE=212, FDR < 0.05) levels undergo rapid changes in response to the mGluR1/5 agonist DHPG. DHPG-responsive RNAs and

RPFs largely did not overlap (n=5). Our DHPG-responsive translational changes in area

CA1 do not overlap with those identified by polysome RNA-seq in cultured cortical neurons, which we attribute to different sample types and methodologies (Di Prisco,

Huang et al. 2014). Gene set enrichment analysis reveals that GO terms related to G protein-coupled receptor activity, synaptic membranes, and the endoplasmic-reticulum membrane are enriched in the upregulated RNA and RPFs, and helicase activity is enriched in the down-regulated RNAs and RPFs (Figure II.9 and Figure II.10 respectively). Some GO terms are unique to the RNA or RPF changes. For example, the

RNAs upregulated following DHPG are enriched for transcription-related functions, and the DHPG translational response is enriched for terms related to the lysosome. Some examples of DHPG-responsive mRNAs and RPFs that are high confidence SFARI autism genes (https://gene.sfari.org, score 1 or 2) are shown in Figure II.8B for Shank3, Dlg4, and Ache. Dlg4 mRNA encodes PSD-95, which has increased mRNA and protein levels following DHPG-mGluR signaling (Zalfa, Eleuteri et al. 2007). Nrsn1 mRNA encodes a neuron-enriched regulator of vesicle transport, which is translationally repressed 5

46 47

minutes after a hippocampus-dependent learning task (Cho, Yu et al. 2015). Translational repression of Nrsn1 mRNA during learning may be a requirement for this cognitive process as evidenced by the observation that its overexpression leads to deficits in object location and contextual fear memory. Because we did not observe immediate early genes

(c-fos, Arc, and Egr1) in our list of differentially expressed RNAs, we infer that the rapid changes in RNA levels are likely due to alterations in RNA destruction and not transcription.

Because FMRP is regulated by and required for mGluR1/5 signaling to translation

(Muddashetty, Nalavadi et al. 2011, Nalavadi, Muddashetty et al. 2012, Niere, Wilkerson et al. 2012, Huang, Ikeuchi et al. 2015), we determined if FMRP mRNA targets are responsive to this form of synaptic plasticity. FMRP binding targets are enriched in the

DHPG-responsive DE RNAs (p < 1.8e-44) and RPFs (p < 5.0e-9). As a group, the FMRP targets have increased RNA levels following mGluR signaling that correlate with stringency of FMRP binding (Figure II.8C, left). FMRP targets have decreased RPFs following synaptic stimulation, and an overall decrease in TE that correlates with FMRP binding stringency (Figure II.8C, middle and right). Therefore, mGluR stimulation regulates RNA levels and translation of FMRP target mRNAs.

47 48

Figure II.8: mGluR signaling induces rapid RNA and RPF changes in wild-type hippocampal slices

A) Heatmaps of mRNAs (left) and RPFs (right) that are responsive to DHPG (100µM, 5min) in wild-type hippocampal slices. Red and blue color represents relative gene expression for individual mRNAs (row); each column represents a biological replicate (n= 7 for RNA-Seq, n= 8 for ribosome profiling), Differentially expressed RNA (DE=241, FDR < 0.01) and RPFs (DE=212, FDR < 0.05). B) Examples of DHPG responsive mRNAs (top) or RPFs (bottom) are shown with normalized RNA and RPF counts from DESeq2. P values are from the output of DESeq2 analysis (Wald-test with BH correction). C) Empirical Cumulative Distribution Function plots (ECDF) of RNA (left), RPF (middle), and TE (right) expression changes in DHPG/Basal (log2FC) of FMRP binding targets compared to non-target mRNAs. FMRP binding targets are stratified into 3 groups based on binding stringency. Wilcoxon rank-sum test with correction by Bonferroni method. p-values are indicated in the plot legend.

48 49

Figure II.9: Gene set enrichment analysis of mGluR-responsive RNAs

Gene set enrichment analysis of GO terms enriched (padj < 0.05) in DHPG/Basal RNAs for molecular function (top) and cellular compartment (bottom). Gene lists were ranked by wild-type DHPG/Basal expression changes (log2FC) and p-value. GO terms are separated by enrichment in either upregulated (increased) or downregulated (decreased) RNAs.

49 50

Figure II.10: Gene set enrichment analysis of mGluR-responsive RPFs

Gene set enrichment analysis of GO terms enriched (padj < 0.05) in DHPG/Basal RPFs for molecular function (top) and cellular compartment (bottom). Gene lists were ranked by wild-type DHPG/Basal expression changes (log2FC) and p-value. GO terms are separated by enrichment in either upregulated (increased) or downregulated (decreased) RPFs.

50 51

mGluR-induced translational dysregulation in TSC Tsc2+/- mice have deficient mGluR-LTD, so we determined whether the RNA and translational response after DHPG is altered. We lacked sufficient statistical power to detect genotype-dependent alterations in the DHPG response in Tsc2+/- slices with their littermate controls. Instead, we compared the genes altered in Tsc2+/- from Figure II.2 and the DHPG-responsive genes in wild-type from Figure II.8 and found a statistically significant overlap for the RNAs (n=23, p < 4.3e-21) and RPFs (n=20, p < 3.8e-18)

(Figure II.11A). Interestingly, 16 of the 23 RNAs that overlap between Tsc2+/- and mGluR1/5-stimulated are FMRP mRNA targets. No GO terms were enriched in the overlapping list, likely due to the small number of genes. Heatmaps of the relative expression levels of the overlapped mRNAs and RPFs in Figure II.11A shows that most of the upregulated mRNAs and RPFs in unstimulated Tsc2+/- slices appear to have a deficient DHPG response compared to wild-type slices Figure II.11B. Our results suggest that some mRNAs and RPFs that are elevated in unstimulated Tsc2+/- slices cannot respond to DHPG.

51 52

Figure II.11: Some mRNAs altered in unstimulated Tsc2+/- slices do not respond to DHPG-mGluR signaling

A) Venn diagrams showing overlap between Tsc2+/- dysregulated RNAs and DHPG- responsive RNAs (top) and Tsc2+/- dysregulated RPFs and DHPG-responsive RPFs (bottom). Gene overlap test was performed with a Fisher’s exact test and a background size of expressed RNAs used in DE analysis. B) Heatmaps of mRNAs (left) and RPFs (right) that overlap from (A). Red and blue color represents relative gene expression for individual mRNAs (row), and each column represents a biological replicate of littermates (n= 3).

52 53

Signaling to eukaryotic elongation factor 2 is altered in the Tsc2+/- hippocampus The sequencing results of the Tsc2+/- slices under basal and DHPG conditions suggest there is a deficiency in the ability of DHPG to regulate translation in the Tsc2+/- mice. To determine whether mGluR1/5 signaling to translational control was altered in

Tsc2+/-, hippocampal slices were treated with DHPG followed by western blotting for several well-established downstream substrates of mGluR activation. Figure II.12A-B shows that under basal conditions, phosphorylated (phospho) ERK or AKT were unaffected by loss of one allele of Tsc2. In response to DHPG treatment, phospho-ERK levels were similarly elevated in WT and Tsc2+/- slices, consistent with other observations (2, 42). DHPG did not induce phospho-Akt in either WT or Tsc2+/- slices.

Because both mTOR and mGluR5 signaling regulate phosphorylation of eukaryotic elongation factor 2 (eEF2), we determined whether signaling to eEF2 could be altered. In wild-type slices, DHPG treatment robustly induces eEF2 phosphorylation (Figure

II.12C). Surprisingly, Tsc2+/- slices have elevated phospho-eEF2 basally, and DHPG treatment fails to induce eEF2 phosphorylation. Total levels of eEF2 are unchanged in

Tsc2+/- at basal (106 ±6% of WT; n = 19 mice; n.s.; 2way ANOVA, posthoc Sidak’s test) and following DHPG treatment (WT 101.7 ±8.02%, Tsc2+/- 114.4 ±9.69%, n=19, n.s.;

2way ANOVA, posthoc Sidak’s test).

53 54

Figure II.12: Deficient mGluR1/5-regulated signaling to translation elongation in Tsc2+/- mice

A-C) The levels of phosphorylated (P) and total (T) ERK (A), AKT (B), and eEF2 (C) in CA1-enriched hippocampal slices from wild-type and Tsc2+/-mouse littermates as determined by western blots under basal or DHPG-treatment condition (100µM; 5min). *p < 0.05, **p< 0.01; ***p< 0.001; 2 way ANOVA, posthoc Sidak’s test, n=11-19 mice/genotype, 4-5 slices per drug condition/mouse. Asterisks below brackets indicate an effect of DHPG on phosphoprotein levels within genotype. Asterisks above brackets indicate effects between genotypes.

54 55

eEF2 is a GTP-dependent elongation factor that promotes transfer of peptidyl- tRNAs from the A to P site of the ribosome. Phosphorylation of eEF2 on threonine 56 prevents it from binding to ribosomes, leading to inhibition of translation elongation

(Carlberg, Nilsson et al. 1990). Somewhat counterintuitively, eEF2 kinase and its phosphorylation of eEF2 are necessary for mGluR-induced synthesis of proteins required for LTD (Park, Park et al. 2008). Global inhibition of translation elongation, such as through phospho-eEF2 or cycloheximide (a chemical inhibitor of ribosome translocation), may promote synthesis of poorly translated mRNAs by liberating rate- limiting translation factors (Park, Park et al. 2008, Acevedo, Hoermann et al. 2018,

Sossin and Costa-Mattioli 2019). Thus, we determined if the mRNAs responsive to mGluR-LTD signaling are enriched for poorly translated mRNAs by using hippocampal translation efficiency as described earlier. We find that the Up RNA group after DHPG treatment has low basal translation efficiency compared to unchanged mRNAs and the

Down mRNA group has high basal translation efficiency (Figure II.13, left). For the translational response, the Up RPF group after DHPG-mGluR signaling has high basal translation efficiency compared to other mRNAs, whereas the Down RPF group had the same translation efficiency as the unchanged mRNAs (Figure II.13, middle). Because

DHPG-mGluR signaling may reduce ribosome translocation via phospho-eEF2, it is possible that at least some of the mRNAs with increased RPFs are associated with stalled ribosomes, in which case they would have decreased translation. Because FMRP binding targets are enriched in the DHPG-mGluR responsive Up mRNA group, we predicted that

FMRP targets are also lowly translated in the hippocampus. Indeed, FMRP binding

55 56

targets have low basal translation efficiency compared to non-targets (Figure II.13, right).

Interestingly, the strength of FMRP binding negatively correlates with basal translation efficiency where the most stringent FMRP targets have lower basal translation efficiency compared to the low and high stringency groups. Taken together, these observations suggest that mGluR5 signaling increases expression of mRNAs that are lowly translated, such as FMRP binding targets.

Figure II.13: Correlation of hippocampal TE with mGluR5-responsive mRNAs and FMRP binding targets

Empirical Cumulative Distribution Function (ECDF) plots of hippocampal TE for the Up and Down RNAs (left) and RPFs (middle) in DHPG/Basal compared to all genes. Right: ECDF plot of hippocampal TE for FMRP binding targets compared to non-target mRNAs. FMRP binding targets are stratified into 3 groups based on binding stringency. Wilcoxon rank-sum test with correction by Bonferroni method. p values are indicated in the plot legend.

56 57

Because Tsc2+/- slices have a deficient phospho-eEF2 response to DHPG treatment, we hypothesized that ribosome elongation is excessive during mGluR signaling in Tsc2+/- and may contribute to deficient mGluR-LTD. We used a low concentration of cycloheximide (75 nM), which inhibits ribosome elongation by binding to the E-site of the ribosome to prevent eEF2-mediated translocation (Schneider-Poetsch,

Ju et al. 2010). This concentration of cycloheximide restored mGluR-LTD in eEF2K knockout mice without affecting mGluR-LTD in WT mice (Park, Park et al. 2008). We induced mGluR-LTD with DHPG in hippocampal slices and tested the effects of low dose cycloheximide (75nM) to inhibit ribosome elongation. While there were no main effects of genotype or treatment overall on LTD magnitude, we observed a significant interaction of genotype*treatment (Figure II.14). With posthoc pairwise comparisons, there was strong trend for cycloheximide to enhance mGluR-LTD in Tsc2+/- (veh: 83±

2% of baseline; cycloheximide; 75± 5%) but not WT mice (veh: 74± 2% of baseline; cycloheximide; 77± 3%). In vehicle, mGluR-LTD in Tsc2+/- was less than WT as previously reported (Auerbach, Osterweil et al. 2011). Our data suggest that the deficiency of mGluR1/5 induced P-eEF2 and suppression of translation elongation in

Tsc2+/- mice limits LTD magnitude and this can be reinstated by pharmacological inhibition of ribosome elongation.

57 58

Figure II.14: Low dose cycloheximide enhances LTD in Tsc2+/-

Left: Time course demonstrating that brief DHPG application (100µM; 5min) induces long-term synaptic depression (LTD) of field (f) EPSP slopes in WT hippocampal slices that is unaffected by low doses of cycloheximide (45-60min pretreatment, 75 nM; n = 12 and 15 slices; 6-7 mice/condition). Middle: In Tsc2+/- mice pre-treatment with low dose cycloheximide enhances LTD magnitude (n= 10 and 15 slices; 7 mice/condition). Plotted are group averages of fEPSP slope (mean ± SEM) normalized to pre-DHPG baseline as a function of time. Inset: Example fEPSP from baseline and during LTD (55-60 min post DHPG). Scale = 0.5mV/10msec. Right: LTD magnitude, measured at 55-60 min post-DHPG application in vehicle and cycloheximide in each genotype. There is a significant interaction of cycloheximide and Tsc2 genotype (F(1,47) = 5.197, p < 0.05; LMM). With posthoc pairwise comparisons, there is a strong trend for cycloheximide to enhance LTD in Tsc2+/- mice (veh: 83± 2% of baseline; cycloheximide; 75± 5%; n = 15 and 10 slices from 4 mice/treatment; F(1,23) = 5.445, # p= 0.058; LMM) but not WT mice (veh: 74± 2% of baseline; cycloheximide; 77± 3%; n = 11 and 15 slices from 6 and 5 mice, respectively/treatment; F(1,24) = 3.819, p= 0.14; LMM). In vehicle, LTD is reduced in Tsc2+/- mice in comparison to WT, but not in low dose cycloheximide. (*p< 0.05; **p< 0.01). Veh = vehicle, CHX = cycloheximide.

58 59

Discussion

Using ribosome profiling and RNA-seq, we identified changes in ribosome footprint number and RNA levels in Tsc2+/- hippocampal slices and following mGluR-

LTD. Our experimental design captures changes from multiple cell types, which is reflective of the tissue complexity of the hippocampus. We found translation of myelin- related mRNAs increased in Tsc2+/- mice, which are predominantly found in oligodendrocytes. The tissue complexity may explain why we did not capture translation of known DHPG-induced mRNAs Arc and Map1b, which are translated following DHPG treatment in dendrites of excitatory neurons (Waung, Pfeiffer et al. 2008, Graber, Hebert-

Seropian et al. 2013).

In complete loss of TSC1 or TSC2 models, there is clearly excessive mTOR activity that leads to widespread changes in translation (Bilanges, Argonza-Barrett et al.

2007). However, mTOR hyperactivation is not always as obvious in Tsc1 and Tsc2 heterozygote models in the brain (Potter, Basu et al. 2013). Using a Tsc2+/- model, we do not observe signs of mTOR hyperactivation by western blot of phosphorylated rpS6 and mTOR in the hippocampus. Regardless, we find evidence of disrupted global mRNA translation by increased TE of TOP mRNAs and elevated levels of phosphorylated eEF2.

The TOP mRNAs, the most well-studied TSC/mTOR translational targets, have increased

TE from predominantly decreased RNA and increased RPF levels. The decreased RNA levels of the TOP mRNAs are surprising, given that one predicts increased expression of

TOP mRNAs from mTOR hyperactivity from loss of TSC2. There could be two possibilities to explain this observation. This may be an example of translational

59 60

buffering to maintain homeostasis of proteins encoded by TOP mRNAs, where RNA levels decrease to offset the increase in ribosome occupancy (McManus, May et al.

2014). Alternatively, the decreased RNA levels of TOP mRNAs may be reflective of the altered signaling to phospho-eEF2, which is increased and opposite of the direction that is predicted by activation of mTORC1-S6K-eEF2K signaling in Tsc2+/- animals (Redpath,

Foulstone et al. 1996, Browne and Proud 2004). mTOR-independent signaling pathways can regulate the TOP mRNAs, such as eEF2K/eEF2 signaling and the RNA-binding protein LARP1, which stabilizes mRNA levels of TOP mRNAs. Decreased ribosomal protein levels can activate a stress response that increases eEF2 phosphorylation

(Gismondi, Caldarola et al. 2014). LARP1 deficiency can also decrease TOP mRNA levels, activate a ribosomal stress response, and increase eEF2K/eEF2 signaling

(Gentilella, Moron-Duran et al. 2017). Consequently, the increased in basal phosphorylated eEF2 in Tsc2+/- mice may occur via other effectors to eEF2K, such as increased intracellular Ca2+ and mGluR1/5 signaling (Wang, Li et al. 2001, Park, Park et al. 2008).

Our finding of increased basal levels of phospho-eEF2 is consistent with the observation of decreased protein synthesis rates in hippocampal slices of Tsc2+/- mice

(Auerbach, Osterweil et al. 2011). eEF2K/eEF2 signaling is required for mGluR-LTD, learning and memory, and maintaining an appropriate excitatory/inhibitory balance in the brain, all of which are disrupted in the Tsc2+/- animals (Sutton, Taylor et al. 2007, Park,

Park et al. 2008, Gildish, Manor et al. 2012, McCamphill, Farah et al. 2015, Weng, Chen et al. 2016, Heise, Taha et al. 2017, Antoine, Langberg et al. 2019). A further direction of

60 61

inquiry is to determine whether genetic or chemical reduction of eEF2K activity can improve molecular, synaptic plasticity, and behavioral phenotypes of the TSC mice.

Our study suggests a more complex role of translational dysregulation in the

Tsc2+/- hippocampus outside of simple mTORC1-rpS6 hyperactivation. Given the many downstream substrates of mTOR, it is possible that some of the mRNA and translational alterations observed in Tsc2+/- are through mTORC1-dependent processes outside of translation (e.g. autophagy, lipogenesis) or mTORC2. Our results also highlight the importance of maintaining appropriate gene dosage in modeling TSC and other mTORopathies such as PTEN in the context of ASD, where only 1 allele is usually mutated in patients.

Tsc2+/- mice have decreased global protein synthesis in the hippocampus, and consistent with this observation, we observed dysregulated eEF2K/eEF2 signaling basally and following activation of DHPG-mGluR signaling in Tsc2+/- mice. Thus, we used the

“ribosome concentration” and eEF2 phosphorylation model to determine if TE correlates with translational changes in Tsc2+/- mice and following DHPG-mGluR5 signaling. The

“ribosome concentration” model was originally proposed to explain the seemingly mRNA-specific effects arising from decreased protein synthesis in ribosomopathies, a group of disorders with mutations in ribosomal proteins that manifest with cell-type specific features. Using mathematical modeling and assuming initiation as the rate- limiting step of translation, the “ribosome concentration” hypothesis proposed that mRNAs with poor initiation rates (low TE) have decreased protein output with reduced global protein synthesis, whereas mRNAs with high initiation rates (high TE) have

61 62

increased protein output. Here, we observe a trend where in Tsc2+/- mice, inefficiently translated mRNAs have decreased translation and efficiently translated mRNAs have increased translation that is consistent with the “ribosome concentration” (Figure II.5).

mGluR-LTD depends on phosphorylation of eEF2, and phosphorylated eEF2 is proposed to increase protein synthesis of poorly translated mRNAs by freeing up rate- limiting translation factors while repressing translation of most mRNAs (Walden and

Thach 1986, Scheetz, Nairn et al. 2000, Park, Park et al. 2008). Surprisingly, during

DHPG-mGluR1/5 signaling, TE correlates more with transcript level changes rather than translation changes, and FMRP binding targets have low TE compared to all genes

(Figure II.13). These data are consistent with the model that phospho-eEF2 increases the abundance of poorly translated mRNAs during mGluR-DHPG signaling, however at the transcript level instead of translation. The link between phospho-eEF2 and regulation of mRNA abundance is unknown, but one possibility is that slowing ribosomes may protect some mRNAs from degradation by decay factors (Chan, Mugler et al. 2018). Another possibility is that phospho-eEF2 is not directly responsible for the mRNA level changes, but instead is part of a parallel mGluR5 signaling pathway that converges on a regulator of mRNA abundance, such as FMRP. Several recent studies have supported a role of

FMRP in modulating stability of its targets through N6-methyladenosine modification of the mRNA and codon optimality (Zhang, Kang et al. 2018, Shu, Donnard et al. 2020). TE is a complex value that includes multiple mRNA features that correlates with efficiency of translation initiation (5′UTR length and structure) and/or mRNA stability (GC content and codon optimality). Thus, the difference in the correlation of TE with the mRNA

62 63

abundance changes following mGluR5 signaling and the translational changes in Tsc2+/- mice may reflect different mechanisms that converge on either mRNA turnover or translation.

RNA-seq reveals that mGluR activation induces a rapid response in both increased and decreased mRNA levels. Because translation is often linked to RNA stabilization/destabilization, this mGluR response could be linked to this process (Richter and Coller 2015). The decrease in specific mRNAs may be the result of heightened rates of degradation, which could be the consequence of mGluR-dependent upregulation of microRNAs (Olde Loohuis, Ba et al. 2015). For mRNAs that increase, there may be a constant rate of mRNA turnover in hippocampal tissue, and mGluR activation could lead to mRNA stabilization through regulation of translation or microRNA dissociation

(Muddashetty, Nalavadi et al. 2011). The DHPG-mGluR responsive mRNAs are highly enriched for FMRP targets and genes encoding synaptic proteins. For example, our study and that of Zalfa et al (Zalfa, Eleuteri et al. 2007) found that levels of PSD-95 mRNA

(Dlg4), an FMRP target, increase after mGluR activation in cultured neurons and is regulated by interaction with a microRNA (Muddashetty, Nalavadi et al. 2011). The observation that DHPG results in decreased RPFs and TE on FMRP target mRNAs suggests they have reduced translation. This would be a surprising finding based on previous findings that DHPG rapidly stimulates the synthesis of proteins encoded by

FMRP target mRNAs such as Map1b, Arc, EF1a and PSD-95 (Waung, Pfeiffer et al.

2008, Waung and Huber 2009, Muddashetty, Nalavadi et al. 2011, Graber, Hebert-

Seropian et al. 2013). Therefore, decreased RPF and TE may not always indicate reduced

63 64

synthesis of protein. Some evidence indicates that FMRP stalls or slows ribosome translocation, and mGluR stimulation regulates translation elongation through FMRP and eEF2 phosphorylation (Park, Park et al. 2008, Darnell, Van Driesche et al. 2011,

Nalavadi, Muddashetty et al. 2012, Shah, Molinaro et al. 2020). Thus, for some mRNAs, the observed decrease in RPF or TE on FMRP target mRNAs in response to DHPG may be a consequence of enhanced ribosome movement that reflects increased protein synthesis.

Recent work has revealed regulation of translational elongation as critical for multiple forms of protein-synthesis dependent synaptic plasticity (Scheetz, Nairn et al.

2000, Park, Park et al. 2008, Verpelli, Piccoli et al. 2010, McCamphill, Farah et al. 2015).

DHPG-induced mGluR1/5 signaling via ERK is normal in the Tsc2+/- hippocampus, which mainly feeds into regulation of translation initiation (Roux and Topisirovic 2018).

Instead, signaling to ribosome translocation (polypeptide elongation) is altered in Tsc2+/- and a low dose cycloheximide enhances mGluR-LTD in these mice. Furthermore, we find evidence for a mechanism of occlusion via eEF2 phosphorylation in Tsc2+/-, where some of the mRNAs and RPFs increased in Tsc2+/- mice appear to no longer respond to

DHPG, consistent with the higher basal levels of phospho-eEF2 and the deficient phospho-eEF2 response following DHPG. Our work suggests a novel mechanism of translational dysregulation in Tsc2+/- and mGluR synaptic plasticity through signaling to ribosome elongation.

A combined Tsc2+/- and Fmr1-/y mouse rescues phenotypes in synaptic plasticity and contextual fear conditioning in each other, yet the mechanism of this rescue paradigm

64 65

has remained elusive. In Tsc2+/- and following DHPG-mGluR signaling, FMRP binding targets have decreased TE from increased RNA levels; in Fmr1-/y, RNA levels of FMRP targets are instead decreased. This bidirectional alteration of FMRP binding targets suggest a mechanism for the opposing changes in synaptic plasticity and rescue paradigm of the Tsc2+/-/Fmr1-/y mice. Because total FMRP protein levels are unchanged in Tsc2+/-, the dysregulation of FMRP binding targets may be through alternate pathways in transcription, mRNA stability, and/or translation. It is unknown if the increased transcript levels of FMRP binding targets are from transcription or post-transcriptional processes, and additional experiments utilizing RNA polymerase II ChIP-seq and metabolic RNA labeling combined with RNA-seq can distinguish contributions from RNA transcription, processing, and turnover. Furthermore, mRNAs that are dysregulated in Tsc2+/- are enriched for DHPG-mGluR responsive mRNAs and many of these overlapping mRNAs are FMRP binding targets. These results reveal a complex interaction of TSC2 and FMRP in steady state and mGluR regulation of distinct mRNA subclasses (Figure II.15).

65 66

Figure II.15: Model of altered eEF2 signaling and FMRP binding targets in Tsc2+/-

Schematic showing DHPG-mGluR5 induction of eEF2K signaling to phosphorylation of eEF2 in hippocampal slices. Left: In WT slices, DHPG-mGluR5 signaling leads to increased levels of phospho-eEF2, which presumably slows ribosome transit and promotes expression of lowly translated mRNAs, such as FMRP binding targets, required for mGluR-LTD. Right: Under basal conditions in Tsc2+/- slices, eEF2 is already phosphorylated by eEF2K, and elevated phospho-eEF2 correlates with increased mRNA abundance of lowly translated mRNAs and FMRP binding targets. DHPG-mGluR5 signaling does not further increase eEF2 phosphorylation in Tsc2+/- slices. Tsc2+/- slices may have excessive ribosome transit during mGluR signaling that contributes to deficient mGluR-LTD.

66 67

Limitations

We anticipated that loss of 1 copy of Tsc2 would result in widespread but perhaps subtle gene expression changes, consistent with other mouse models of autism. As such, we used a p-adjusted cut-off of 0.1 for the genotype comparison. One limitation of the interpretation of our ribosome profiling results is that we identified altered signaling to ribosome movement in Tsc2+/- and following mGluR-LTD signaling. Thus, changes in ribosome density may not a priori reflect changes in protein synthesis. Multiple mouse models of autism have altered mGluR-LTD signaling, but evidence of altered mGluR-

LTD in humans with autism has yet to be established.

Conclusions

By performing RNA-seq and ribosome profiling on a mouse model of Tsc2+/- autism and mGluR-LTD, we identified changes in FMRP binding targets that suggest a molecular basis for bidirectional regulation of synaptic plasticity and behavior by TSC2 and FMRP. Our study also suggests that altered mGluR-regulated translation elongation contributes to impaired synaptic plasticity in Tsc2+/- mice.

Material and methods

Animals

Germline Tsc2+/- mice were generated by crossing CMV-Cre (B6.C-Tg(CMV- cre)1Cgn/J) (JAX stock number: 006054) with Tsc2flox mice (JAX Stock number:027458). F1 were then bred on the C57BL/6J background to obtain WT and

67 68

Tsc2+/- mice. Animals were given ad libitum access to food and water and reared on a 12- hour light-dark cycle. All experiments were approved by the Institutional Animal Care and Use Committee at University of Texas Southwestern and conducted in accordance with the National Institutes of Health Principles of Laboratory Animal Care.

Hippocampal slice preparation

Acute hippocampal brain slices were prepared from P30-40 wild-type (WT) and Tsc2+/- littermates as described previously (Jakkamsetti, Tsai et al. 2013). Briefly, mice were anesthetized with ketamine (125 mg/kg)/xylazine (25 mg/kg) and transcardially perfused with chilled (4°C) sucrose dissection buffer containing the following (in mm): 2.6 KCl,

1.25 NaH2PO4, 26 NaHCO3, 0.5 CaCl2, 5 MgCl2, 212 sucrose, and 10 dextrose aerated with 95% O2/5% CO2. Transverse hippocampal slices (400µm) were obtained on a Leica

VT1200S slicer. CA3 was cut off to avoid epileptogenic activity induced by DHPG.

Slices were recovered for 3-4 hours and maintained at 30°C in artificial cerebral spinal fluid (ACSF) containing the following (in mM): 119 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2,

26 NaHCO3, 1 NaH2PO4 and 11 D-glucose aerated with 95% O2/5% CO2 to pH 7.4.

Slices were then treated with cycloheximide (100 µg/ml) in ACSF, 4°C, snap-frozen on dry ice/EtOH bath, and stored at -80°C. Alternatively, hippocampal slices were treated with DHPG (100 µM) (Tocris) or artificial cerebrospinal fluid (ACSF, vehicle) for 5 min.

For the cycloheximide experiments, hippocampal slices were pre-treated with 75nM cycloheximide for 45-60 minutes. At the end of the incubation, CA1 was dissected out, snap frozen in a dry ice/ethanol bath, and stored at -80 C. Some hippocampal slices were

68 69

prepared without removing CA3 and used for RNA-seq and ribosome profiling experiments as indicated below (batch 1-5).

Electrophysiology recordings

For all recordings, slices were submerged and perfused with ACSF at 2.5–3.5 ml/min (30

± 1°C). Field potentials (FPs) and EPSCs were evoked by stimulation of the Schaffer collateral pathway with a concentric bipolar tungsten electrode. Field potentials were recorded with a glass electrode (1 MΩ) filled with ACSF placed in the stratum radiatum of CA1. Test stimuli were delivered every 30 s and a stable baseline was obtained at

∼50% of the maximum FP amplitude.

Western Blotting

Western blotting was performed on whole hippocampus homogenates or on CA1 hippocampal slices, where indicated, and as previously described (Ronesi, Collins et al.

2012). CA1 hippocampal slices were lysed with lysis buffer (50 mm Tris, pH 7.4, 120 mm NaCl, 50 mm NaF, and 1% Triton X-100, containing Protease Inhibitor Mixture,

Sigma #P8340; and phosphatase inhibitor mixture 2 and 3, Sigma #P5726 and #P0040).

Samples were homogenized using brief (5–10 s) pulses of sonication with an ultrasonic cell disruptor until lysates were clear. Lysates were then centrifuged at 15,700 ×g for 10 min at 4°C, the pellets discarded, and protein levels in the supernatant were measured using a Pierce BCA kit. After SDS-PAGE, proteins were transferred onto PVDF membranes. After blocking with 5% BSA in 1× TBS, 0.05% Tween 20 for 1 h,

69 70

membranes were incubated with the following primary antibodies in blocking buffer overnight at 4°C: The following antibodies were used in this study: S6 Ribosomal Protein

(Cell Signaling, #2217), Phospho-S6 Ribosomal Protein Ser235/236 (Cell Signaling,

#4856), Akt (Cell Signaling, #9272), Phospho-Akt Ser473 (Cell Signaling, #4060), p44/42 MAPK (Erk1/2) (Cell Signaling #9102), Phospho-p44/42 MAPK (Erk1/2)

Thr202/Tyr204 (Cell Signaling, #9101), eEF2 (Cell Signaling, #2332), Phospho-eEF2

Thr56 (Cell Signaling, #2331), mTOR (Cell Signaling, #2983), Phospho-mTOR Ser2448

(Cell Signaling, #2971), FMRP (DSHB, 2F5-1), Glyceraldehyde-3-Phosphate

Dehydrogenase (Millipore, #MAB374), actin (Millipore, #MAB1501). After three 10min washes of 1× TBS, 0.05% Tween 20, membranes were incubated with the appropriate

HRP-conjugated secondary antibodies in 5% milk in 1× TBS, 0.05% Tween 20 for 1 hr at room temperature. After three 10min washes of 1× TBS, 0.05% Tween 20, the membranes were developed using ECL. For comparison of phosphorylated protein levels across conditions or genotypes, immunoreactive phosphorylated protein bands were normalized to total protein levels from the same slice homogenates (e.g., P-ERK/ERK), each of which was first normalized to loading control (either GAPDH or actin where indicated).

Sample preparation for RiboSeq and RNA-Seq

Hippocampal slices from 6 mice were pooled, lysed in polysome buffer (as above but with the addition of 100ug/mL cycloheximide, 25U/mL Turbo Dnase I, 1X EDTA-free protease inhibitor, 1% NP-40), and triturated with a 25G needle 10 times. Following

70 71

placement on ice for 10 minutes, the lysate was centrifuged 2,000g for 10min at 4°C, and the supernatant collected and centrifuged again for a 20,000g for 10min at 4°C. This supernatant was collected and the RNA concentration measured using the Qubit RNA HS

Assay kit (Invitrogen, #Q32852). One-fourth of the lysate was saved for RNA-seq libraries (~1ug of RNA) and the remainder (~3.5ug) was treated with 11.3ug of RNase A

(Ambion, #AM2270) and 1410U of Rnase T1 (Thermo Fisher Scientific, #EN0542), which corresponds to a ratio of ~3ug Rnase A/375U Rnase T1 per ug of input RNA. The

Rnase treatment was carried out at 25°C for 30min with gentle shaking and the digest stopped by placement on ice and the addition of 2.5uL of SuperRNaseIn (Invitrogen,

#AM2696). The lysate was layered over a linear 10-50% sucrose gradient (prepared with polysome buffer supplemented with 1mM DTT and 100ug/mL CHX) and centrifuged at

35,000rpm for 2.5 hours at 4°C in a SW41 rotor. The 80S monosome fractions were identified by continuous A260 monitoring and pooled for RNA extraction by Trizol LS.

To aid in RNA recovery of low concentration samples, 20ug glycogen was added to all precipitation steps. A total of 3 biological replicates per genotype and treatment were made from hippocampal slices (CA1 enriched, batch 6-8) from Tsc2+/- and their wild- type littermate mice. Wild-type slices had a total of 8 biological replicates for RiboSeq and 7 for RNA-Seq (batch 1-5 from hippocampal slices with CA3, batch 6-8 from CA1 enriched).

71 72

Library Preparation

All RiboSeq libraries were prepared as previously described (Heyer, Ozadam et al. 2015).

RiboZero Gold (Illumina, #MRZG12324) was used to deplete rRNA for RiboSeq libraries. The purified 80S monosome mRNA fragments were resolved on a 15% Tris- borate urea gel (National Diagnostics, #EC-833) and ribosome protected fragments between 26nt and 34nt were recovered from the gel by the crush-soak method overnight in RNA extraction buffer (300mM NaOAc pH5.5 1mM EDTA 0.25% SDS) at 25°C with constant horizontal rotation. The RNA was then precipitated and the purified RNA was dephosphorylated with PNK enzyme (NEB, #M0201S) and ligated to preadenylated adaptors (IDT) using T4RNL2Tr.K227Q ligase (NEB, #M0351L). Reverse transcription

(RT) was performed using SuperScript III (Invitrogen, #18080-044) in a modified 1X

First Strand synthesis buffer without MgCl2 (50 mM Tris–HCl, pH 8.3, 75 mM KCl).

Primers containing a 5nt barcode and 8nt unique molecular identifier (UMI) were used for RT. After RT, the RNA was hydrolyzed with a final concentration of 0.1N NaOH for

20min at 98°C. The cDNA was resolved on a 10% TBU gel and the 130-140nt band was selected. The cDNA was eluted from crushed gel pieces overnight in DNA extraction buffer (300mM NaCl, 1mM EDTA, 10mM Tris pH8.0) at 25°C with constant horizontal rotation. The cDNA was circularized using CircLigase (Epicentre, #CL4115K) and depleted of tRNA and rRNA using biotinylated anti-sense probes (IDT) and Dynabeads

MyOne Streptavidin C1 (Invitrogen, #65001). To determine the optimal number of PCR cycles, a test PCR was performed using the KAPA library amplification kit (Kapa

Biosystems, #KK2611). The final PCR product was run on an 8% TBE gel, and the 180-

72 73

190nt bands were selected. The PCR product was eluted overnight in DNA extraction buffer as described earlier.

Batch 1-5 RNA-Seq libraries were prepared using RiboZero to deplete rRNA and then fragmented with 2X alkaline fragmentation solution (2 mM EDTA, 10 mM Na2CO3,

90 mM NaHCO3, pH ~ 9.3) at 95°C for 15min to yield ~140nt fragments. The fragmented RNA was separated on a 10% TBU gel, and those migrating between 100 and

150nt were selected. RT products were separated on a 10% TBU gel and the 200-250nt band was selected. The final PCR product was run on an 8% TBE gel, and the 260-300bp band was selected. The size distribution of the final libraries was confirmed by Fragment

Analyzer (UMMS Molecular Biology Core). Batch 6-8 RNA-Seq libraries were prepared using NEXTflex poly(A) beads (Bioo Scientific, #512980) and the NEXTFlex Rapid

Directional qRNA-Seq kit (Bioo Scientific, # NOVA-5130-03D) according to manufacturer’s instructions.

The final libraries were purified using AMPure XP beads (Beckman Coulter,

#A63880) Purified libraries were quantified by qPCR using KAPA Library

Quantification kit (Kapa Biosystems, #KK4835) and were pooled to equimolar ratios.

Libraries were sequenced on a NextSeq 500/550 High Output Kit v2 (Illumina, #FC-404-

2005) as 75bp single-end runs on a NextSeq500 sequencer.

RiboSeq pipeline and read mapping

The individual samples were separated based on 8nt barcodes using custom scripts.

Cutadapt (1.7.1) was used to remove the adapter

73 74

(TGGAATTCTCGGGTGCCAAGGAGATCGGAAGAGCGGTTCAGCAGGAATGCC

GAGACCG). The reads were then uploaded to the UMass Bioinformatics Core Dolphin platform for mapping (https://www.umassmed.edu/biocore/introducing-dolphin/). The reads were quality filtered using Trimmomatic (0.32) and then mapped to the rRNA and tRNA reference using bowtie2 (2.1.0). The rRNA and tRNA unmapped reads were kept and then mapped to mm10 using TopHat2 (2.0.9). Only uniquely mapped reads were kept, and PCR duplicates were marked based on UMI sequence and a custom script. samtools (0.019) was used to keep uniquely mapped reads without duplicates. For gene level quantification of RiboSeq, RSEM/1.2.11 was used to align cleaned, unique reads to the RefSeq mm10 CDS with the first and last 30nts removed to avoid quantifying initiation and termination peaks. For RNA-seq libraries, reads were aligned to the entire mm10 transcriptome. Counts from RSEM were batch corrected using the Combat function in sva (3.24.4). Batch-corrected counts were used for DESeq2 (1.26.0) to identify differentially expressed genes for RNA and RiboSeq libraries. Normalized RNA and RPF counts from DESeq2 were used for generating plots. To generate relative expression heatmaps of differentially expressed genes, regularized log2 transformation was applied to read counts for variance stabilization. The row mean was subtracted from individual row value to yield relative expression level of each gene. anota2seq (1.0.1) was used to calculate expression changes in TE (log2FC) using the Relative Log

Expression (RLE) for normalization.

74 75

GO term, GSEA, and gene overlap analysis

GO term enrichment was performed using the R package clusterProfiler (3.14.0)(Yu,

Wang et al. 2012). For GSEA, mRNAs were ranked by their log2FoldChange and pvalue from DESeq2 output. P values were corrected with Benjamini-Hochberg, and terms with padj < 0.05 were considered enriched. Gene overlap analysis was performed using the R package GeneOverlap (1.22.0). The genes past filtering from DE analysis were used as the background.

Transcript feature analysis

Gene features were obtained from the UCSC Genome Browser using RefSeq annotations.

For downstream transcript analysis, the most abundant isoform from the RSEM output of the RNA-Seq libraries was selected. ViennaRNA/RNAFold (2.1.6h) was used to calculate the minimal free energy of the 5′UTR.

Hippocampal TE calculation and rolling mean plots

The average CA1 translational efficiency (TE) was calculated by taking the ratio of RPF to RNA reads per kilo-base per million mapped reads (RPKM) from 3 wild-type basal samples aligned to the mm10 CDS reference without the initiation and termination peak.

For consistent TE calculations, mapping to the CDS for RPF and RNA was used, and transcripts longer than 300nt and RPKM > 5 were included.

75 76

Quantification and statistical analysis

An unpaired Wilcoxon rank-sum test with Bonferroni correction was used to compare means between groups unless indicated otherwise in the text and figures. A one-tailed

Kolmogorov-Smirnov test was used in Figure II.5. An unpaired t-test was used for quantification of western blots in Figure II.4B and Figure II.7C. Western blot results in

Figure II.12A-C were analyzed with a 2 way, repeated measures ANOVA. Factors are

DHPG treatment and genotype, repeated measures are basal and DHPG levels in the same mouse. n= # mice. Posthoc Sidak’s multiple comparison test was performed to determine statistical significance for specific comparisons, reported in the figure. For analysis of LTD results in Figure II.14, we performed a linear mixed-effects model

(LMM) statistic, whereby the variability due to random effects (animal, slice) was taken into account, allowing for direct measurement of genotype and/or treatment effects

(Nakagawa and Hauber 2011). Mixed-effects models were fitted using IBM SPSS package. Fixed factors were genotype, treatment, genotype*treatment; repeated/random effects were slices nested within subject (mouse).

Code availability

Custom scripts for generating plots will be available upon request to the lead contact.

Availability of data and materials

The RNA-seq and Ribo-seq datasets supporting the conclusions of this article are available in the GEO Database repository: GSE144539. Publically available datasets

76 77

analyzed in the current study are from the following GEO Database repository: Shah et al, 2020: GSE143333

Acknowledgements

We would like to thank members of the Richter lab for helpful comments, Elisa Donnard for setting up the RiboSeq pipeline, and Ruijia Wang, Julie Zhu, and Jeon Lee for help with the bioinformatics analysis.

77 78

CHAPTER III GLOBAL PROFILING OF TRANSLATION BY FMRP-ASSOCIATED RIBOSOMES

Annie Hien and Joel D. Richter

Program in Molecular Medicine

University of Massachusetts Medical School

Worcester, MA 01605

78 79

Preface

A.H. and J.D.R. designed the experiments. A.H. performed all the experiments, analysis, and writing with input from J.D.R.

Introduction

Fragile X Related Protein (FMRP) is a RNA-binding protein (RBP) whose loss causes Fragile X Syndrome, the leading monogenetic cause of intellectual disability.

Multiple studies have shown that the vast majority of FMRP interacts with ribosomes, possibly to slow ribosome movement on mRNA to repress translation. Thus the FMRP- ribosome complex is likely critical for translational repression of mRNAs linked to synaptic function and autism, yet their identities are largely unknown. This project seeks to identify these mRNAs using selective ribosome profiling.

Fragile X Mental Retardation Protein (FMRP) is an RNA-binding protein (RBP) with roles in mRNA stability, localization, and translational repression of targets important for neurodevelopment. FMRP contains 2 canonical RNA-binding domains, the

K homology (KH) domains and the arginine and glycine rich (RGG) box. Cross-linking immunoprecipitation (CLIP) experiments to identify direct targets of FMRP revealed a surprising binding pattern for FMRP. Unlike most RBPs that bind a specific motif in the untranslated regions (UTRs), FMRP exhibits broad binding to the CDS of roughly 1000 mRNAs linked to synaptic function and autism (Darnell, Van Driesche et al. 2011, Tabet,

Moutin et al. 2016, Maurin, Lebrigand et al. 2018, Sawicka, Hale et al. 2019, Li, Shin et al. 2020). CLIP experiments find that FMRP binding targets are enriched for ACUK and

79 80

WGGA motifs and G-quadruplex containing mRNAs (secondary structures that form in guanine-rich mRNAs). (Ascano, Mukherjee et al. 2012, Maurin, Lebrigand et al. 2018,

Li, Shin et al. 2020). However, the ACUK and WGGA motifs are found in all mRNAs, so it is unclear how FMRP mediates specificity through these sequences. There is also little consensus as to whether these cis-elements are important for FMRP regulation since the ACUK/WGGA sequence and G-quadruplex structure had no effect on FMRP translational inhibition in vitro (Chen, Sharma et al. 2014).

A possible explanation of FMRP’s binding preference for CDS and mechanism of translational repression is through an interaction with the ribosome. Most of FMRP associates with polyribosomes (Khandjian, Corbin et al. 1996, Feng, Warren et al. 1997,

Stefani 2004), possibly as part of a stalled ribosome-mRNA complex (Darnell, Van

Driesche et al. 2011). FMRP-bound ribosome complexes are resistant to puromycin, an aminoacyl-tRNA analogue that promotes premature chain termination of elongating ribosomes, suggesting that FMRP-associated ribosomes are not in an elongation state or that FMRP prevents the action of puromycin. After ribosome runoff, FMRP binding targets co-sediment in lighter polysome fractions in multiple FMRP loss of function models (Darnell, Van Driesche et al. 2011). In addition, ribosome transit rate is faster in

Fmr1-/y hippocampal slices (Udagawa, Farny et al. 2013). Collectively, these experiments suggest that FMRP can block ribosome translocation to repress translation of its target mRNAs. Multiple studies have suggested that FMRP can directly associate with ribosomes via the 60S subunit (Khandjian, Corbin et al. 1996, Siomi, Zhang et al. 1996,

Ishizuka, Siomi et al. 2002). In vitro binding assays demonstrates that exons 13 and 14 of

80 81

human FMRP can directly interact with the 60S in the absence of mRNA (Siomi, Zhang et al. 1996). Exons 13 and 14 are predicted to form a coiled coil structured, which mediates protein-protein interaction. In addition, a study found that recombinant drosophila FMRP (dFMRP) can directly bind the ribosome with high affinity (KD ~

20nM) to repress translation in vitro (although this study concludes that the N-terminal

KH domains mediate FMRP’s interaction with the ribosome) (Chen, Sharma et al. 2014).

Chen et al solved a cryo-EM structure of dFMRP directly bound to the ribosome, which indicates that dFMRP binds the ribosome intersubunit space to occlude P-site aminoacyl- tRNAs and elongation factors from binding. However, it is unclear how a translationally- repressed ribosome can accommodate both FMRP and tRNA in the P-site. The authors proposed a model where the RNA-binding properties of the RGG box of FMRP mediate the mRNA-specificity of the FMRP-ribosome complex.

The cryo-EM structure of FMRP and the ribosomes raises an interesting question of whether this complex can recognize specific mRNAs similar to or distinct from FMRP binding targets. FMRP has numerous functions, and while CLIP studies have identified

FMRP binding targets, the functional consequences of FMRP binding to mRNA cannot be inferred by simply identifying binding partners. In addition, CLIP experiments may miss certain populations of FMRP targets: UV crosslinking efficiency is poor and FMRP-

CLIP targets are biased to identifying long, highly abundant transcripts (Ouwenga 2015).

Conventional ribosome profiling can be used to identify translational dysregulation in

Fmr1-/y mice, a mouse model of FXS. However, FMRP binds roughly 5% of brain mRNAs with diverse functions. Thus, conventional ribosome profiling makes it

81 82

challenging to distinguish direct effects from secondary effects arising from loss of

FMRP. Selective ribosome profiling (SeRP) of an enriched FMRP-containing ribosome population can address this issue by identifying direct targets of FMRP-ribosomes that are stalled (Figure III.1) (Oh, Becker et al. 2011, Becker, Oh et al. 2013). FMRP is particularly amenable to SeRP based on the current model that 1) FMRP associates with polysomes (possibly through a direct interaction), 2) FMRP stalls ribosome movement, and 3) FMRP regulates translation and mRNA stability (Darnell, Van Driesche et al.

2011, Chen, Sharma et al. 2014). Furthermore, FMRP selective ribosome profiling may enrich for FMRP targets that are lowly expressed and missed by CLIP studies. Thus, selective ribosome profiling may identify the mRNAs bound by FMRP-associated ribosomes and could reveal insight into how FMRP-ribosome complexes recognize and regulate its targets.

82 83

Figure III.1: Comparison of conventional and selective ribosome profiling

Schematic of conventional ribosome profiling and selective ribosome profiling. Both methods use RNase to digest the mRNA to ~30 nucleotide fragments, termed ribosome protected fragment (RPF). The RPFs are purified either by sucrose cushion or gradient for conventional ribosome profiling or affinity-purification for selective ribosome profiling. The RPFs are collected for library preparation, high- throughput sequencing, and mapping to the transcriptome.

83 84

Results

FMRP interacts with ribosomal proteins after RNase digestion Ribosome profiling requires digestion of mRNA with RNase, leaving a ~30 nucleotide ribosome protected fragmented (RPF). Conventional ribosome profiling captures all ribosomes, whereas selective ribosome profiling (SeRP) enriches for a ribosome-population of interest, typically through affinity purification. Because

FMRP is an RNA-binding protein, we first determined whether FMRP can still associate with ribosomes after RNase treatment. We added RNase A up to the concentration normally used for ribosome profiling experiments, applied the lysate to a sucrose cushion, and observed where proteins found after centrifugation

(Figure III.2, left). Protein complexes sediment primarily based on viscosity, molecular composition, and weight; ribosomes are large, complex, and heavy structures and readily pellet. Figure III.2 shows that ribosomal proteins RPS6 and

RPL4 are found in the pellet at all RNase concentrations as expected. FMRP is also predominantly found in the pellet whereas HuR, a RBP that associates with ribosomes in a RNA-dependent manner, is released from the pellet to the supernatant at the highest RNase A concentration. This result suggests that after digesting lysates with RNase A at a concentration of 100ng/A260, FMRP co- sediments with ribosomal proteins whereas the RBP HuR is released to the supernatant.

84 85

Figure III.2: FMRP co-sediments with ribosomal proteins independent of RNA

Left: Schematic showing RNA-dependent and RNA-independent mechanisms of how RNA-binding proteins (RPBs) FMRP and HuR interact with ribosomes. FMRP preferentially binds to CDS regions and HuR binds specific motifs in the 3′ untranslated region (UTR) (Darnell, Van Driesche et al. 2011). RNase is used to digest RNA, freeing RBPs that interact with ribosomes in an RNA-dependent manner. Applying the lysate through a sucrose cushion allows for proteins to sediment based on size, with lighter complex at the top and heavier complexes at the bottom such as ribosomes. After RNAse treatment, RBPs that directly bind the ribosomes will remain in the ribosome pellet. Right: Western blot of ribosomal proteins, FMRP and HuR from the supernatant and pellet of the sucrose cushion with increasing amounts of RNase A. S: supernatant, P: pellet

85 86

Co-sedimentation of factors does not necessarily reflect a direct interaction; consequently, we performed immunoprecipitation of HEK cells and cortical mouse tissue after RNase treatment. RPL4 is a core component of the 60S subunit that is found in all ribosomes. Figure III.3 shows that immunoprecipitation with RPL4 antibody co-immunoprecipitates (co-IPs) ribosomal proteins RPLP0 and RPS6 as expected, because they are all constituents of the ribosome. Immunoprecipitation with RPL4 antibody also co-IPs FMRP, indicating that FMRP and the ribosome are likely part of the same complex. The reverse immunoprecipitation with FMRP antibody shows that endogenous FMRP co-IPs RPL4, RPLP0, and RPS6. Negative controls HuR and FUS are RBPs that do not interact directly with ribosomes and are not co-IP’d by the RPL4 nor FMRP immunoprecipitation. The IgG negative control immunoprecipitation pulls down very little of the probed proteins. Finally, immunoprecipitation with FMRP or RPL4 antibodies in the cortex yields similar results as HEK cells (Figure III.3, right). Collectively, these results suggest that FMRP interacts with 40S and 60S ribosomal proteins after RNase digestion of free mRNA, supporting the model that FMRP can directly interact with the ribosome.

86 87

Figure III.3: FMRP interacts with ribosomal proteins independent of RNA

Western blots demonstrating that FMRP can interact with ribosomal proteins independent of RNA in HEK cells (left) and mouse cortex (right). Immunoprecipitation was performed with RPL4, FMRP, or IgG negative control as indicated on the top. Proteins probed by western blot are indicated on the left of the blots. Negative controls RBP HuR and FUS do not interact with ribosomal proteins (RPL4, RPLP0, RPS6) after RNase treatment. TUB: tubulin.

87 88

Anti-FMRP enriches for specific RPFs over total ribosomes After determining the feasibility of selective ribosome profiling of FMRP- associated ribosomes, RNA was extracted from the immunoprecipitation conditions described earlier. RNA from the following IPs were prepared and used for the following comparisons in Table III-1.

Table III-1: List of sample types and purpose for FMRP selective ribosome profiling

Sample name Purpose

Wild-type FMRP-IP FMRP-enriched ribosome pool Total ribosome pool as a reference to calculate FMRP- Wild-type RPL4-IP ribosome enrichment Negative control to establish background, nonspecific Wild-type IgG-IP binding to beads Fmr1-/y FMRP-IP Negative control to establish FMRP antibody specificity

Identifies changes in ribosome occupancy between wild- Fmr1-/y RPL4-IP type and Fmr1-/y enriched mRNAs

Figure III.4 shows that the majority of reads from selective ribosome profiling mapped to protein-coding regions, as expected for ribosome profiling libraries. Roughly 20% of reads mapped to the 5′ and 3′UTR, comparable to conventional ribosome profiling libraries. Figure III.5 shows a volcano plot with 125 mRNAs that are enriched in FMRP/co-IP’d ribosomes over total L4/co-IP’d ribosomes and IgG negative control (log2FC > 0.75, padj < 0.01) (see Table III-2 for full list). The mRNAs enriched by anti-FMRP ribosomes overlapped significantly

88 89

with FMRP targets identified by CLIP experiments in the cortex (25/125) and SFARI

autism genes (23/125). However, many mRNAs did not overlap with direct FMRP

binding targets, suggestive of a novel mode of regulation by FMRP.

Figure III.4: Gene body mapping frequency for conventional and selective ribosome profiling

Percentage of uniquely mapped reads to gene regions from conventional and selective ribosome profiling from RPL4-IP and FMRP-IP from the cortex. Blue = CDS, red = 5′ and 3′ untranslated regions (UTRs), green = introns, purple = intergenic.

89 90

Figure III.5: Anti-FMRP enriches for specific mRNAs

Volcano plot of RPFs that are enriched in anti-FMRP ribosomes over total ribosomes and IgG negative control (x-axis: log2 fold enrichment, y-axis: p- adjusted value with BH correction). 125 mRNAs pass a statistical cutoff of padj < 0.01 and log2FC > 0.75. Venn diagram showing gene overlap of the 125 mRNAs enriched in anti-FMRP ribosomes compared to FMRP targets identified by CLIP in the mouse cortex (Darnell, Van Driesche et al. 2011) and SFARI autism genes (gene.sfari.org).

90 91

Figure III.6 shows examples of mRNAs enriched in anti-FMRP-associated ribosomes. Nuclear receptor corepressor 1 (Ncor1) and E1A binding protein p300

(Ep300) are consistent FMRP targets identified from 3 CLIP studies and show an even distribution of anti-FMRP ribosome densities along the CDS (Figure III.6, left)

(Darnell, Van Driesche et al. 2011, Ascano, Mukherjee et al. 2012, Maurin, Lebrigand et al. 2018). Both have recently been implicated in epigenetic alterations in mice that contribute to FXS phenotypes (Darnell, Van Driesche et al. 2011, Ascano, Mukherjee et al. 2012, Korb, Herre et al. 2017, Li, Stockton et al. 2018, Maurin, Lebrigand et al. 2018).

For some mRNAs, anti-FMRP ribosomes have a unique binding pattern compared to total ribosomes. For example, lysine demethylase 5C (Kdm5c ) and forkhead box P1

(Foxp1) have enrichment of anti-FMRP ribosomes in the 3′ end of the CDS, which may be suggestive of stalled ribosomes or recognition of a specific section of mRNA

(Figure III.6, right). Both Kdm5c and Foxp1 are high confidence SFARI autism genes with mutations identified in humans with intellectual disability (gene.sfari.org). In summary, FMRP 1) co-IPs ribosome complexes and identifies specific RPFs, 2) are enriched over total ribosomes and IgG negative control, and 3) have unique binding preference on CDS.

91 92

Figure III.6: Different modes of binding by anti-FMRP ribosomes

Read density plots for mRNAs enriched by anti-FMRP ribosomes. Normalized reads are plotted along the CDS position in nucleotides. Solid red line = WT FMRP-IP, dashed red line = Fmr1-/y FMRP-IP, solid blue line = WT RPL4-IP, and dashed blue line = Fmr1-/y RPL4-IP.

92 93

Anti-FMRP enriches for specific RPFs independent of FMRP When we compared the FMRP coIP-enriched mRNAs between wild-type and

Fmr1-/y mice, we were surprised to see very few differences between the two samples (Figure III.7A). The few mRNAs that were differentially enriched in the

FMRP co-IP’d ribosomes had differences in total RPFs (e.g. Fmr1) that reflect differences in either total RNA or translation. Given that the reads enriched by anti-

FMRP likely reflect real ribosome fragments, there could be 3 possibilities to explain these results. Figure III.7B depicts a model where anti-FMRP could recognize i) another ribosome-associated protein, ii) the growing nascent peptide on the ribosome, or iii) a co-translational interactor that binds the growing nascent peptide. For the first possibility, the most logical candidates would be the autosomal paralogues, fragile X related proteins 1 and 2 (FXR1/2P). In mammals, FXR2P and

FXR1P share a high level of homology with FMRP. FXR2P and FXR1P are also found in polyribosomes, and the FMRP region required for binding to the 60S (exon 13 and

14) is conserved with dFMRP and FXR2P (Siomi, Zhang et al. 1996, Darnell, Van

Driesche et al. 2011). In drosophila, there is one FMRP that is structurally more similar to the FXRs than to mammalian FMRP (Figure III.7C). Mammalian FMRP contains a unique expansion of exon 11 and 12 and a shorter C-terminus compared to dFMRP and FXR2P.

93 94

Figure III.7: Anti-FMRP enriches for specific mRNAs independent of FMRP

A) Scatter plot of counts from FMRP-IP of wild-type and Fmr1-/y cortices shows very few mRNAs differentially enriched based on genotype. mRNAs that are differentially enriched in the FMRP-IP are from changes in total ribosomes (e.g. Fmr1). B) A model of how anti-FMRP may enrich for specific RPFs via recognition of i) another ribosome-associated protein (e.g. paralogue FXR2P), ii) the nascent peptide of the translated mRNA, or iii) through a co-translational interactor. C) The Fragile X related family of proteins have conserved protein domains. The agenet- like domains are required for protein-protein interactions, such as with FXR1P/2P, CYFIP1/2 and NUFIP. The K Homology (KH) domains and arginine-glycine-rich (RGG) box mediate the RNA binding function of FMRP. The KH domains are required for FMRP association with polyribosomes. The RGG box mediates FMRP binding to G-quadruplex containing mRNAs. The asterisk and red box corresponds to the region required for 60S binding. The solid line indicates the FMRP-specific expansion in exon 11 and 12.

94 95

We observed that Fmr1 and Fxr2 mRNAs were enriched in the anti-FMRP ribosome list from Figure III.5/Table III-2 and proposed that anti-FMRP recognized the growing nascent peptide from Fmr1 and Fxr2 translation. Normalized read density plots for Fmr1 and Fxr2 are indicated in Figure III.8A. For Fxr2, the peak enrichment of anti-FMRP ribosomes correlate with the emergence of the FMRP antibody epitope (amino acids 550-end of human FMRP based on the manufacturer’s datasheet) from the ribosome, suggesting that anti-FMRP was enriching for Fxr2 footprints by cross reacting to the epitope of the nascent chain.

Interestingly, Fmr1 has a different binding pattern from Fxr2, where the peak enrichment of anti-FMRP ribosome footprints occurs before the epitope region. This suggests that anti-FMRP binds predominantly full-length FMRP, and FMRP- associated ribosomes may specifically recognize Fmr1 directly or via a co- translational interactor (Figure III.7B, model I and III). Immunoprecipitation using anti-FMRP in wild-type and Fmr1-KO cortex reveals that anti-FMRP pulls down

FXR2P in Fmr1-/y animals, confirming that the antibody cross-reacts and pulls down

FXR2P (Figure III.8B). Functionally, binding targets of FXR2P and FXR1P overlap

95% with FMRP targets identified by CLIP in HEK cells (Ascano, Mukherjee et al.

2012). The mRNAs enriched in anti-FMRP overlap significantly with FXR2P binding targets in HEK cells (Figure III.8C). Collectively, these results suggest that some mRNAs enriched by anti-FMRP may occur by cross-reactivity to the FXR2P.

95 96

Figure III.8: Anti-FMRP cross-reacts to paralogue FXR2P

A) Read density plots for Fxr2 (top) and Fmr1 (bottom). Normalized reads are plotted along the CDS position in nucleotides. Solid red line = WT FMRP-IP, dashed red line = Fmr1-/y FMRP-IP, solid blue line = WT RPL4-IP, and dashed blue line = Fmr1-/y RPL4-IP. The gray box corresponds to the FMRP antibody epitope. B) Western blot showing that immunoprecipitation with FMRP antibody pulls down FXR2P in wild-type and Fmr1-/y cortex. C) Overlap of anti-FMRP ribosome enriched mRNAs and FXR2P CLIP targets from HEK cells (Ascano, Mukherjee et al. 2012). P-value from Fisher exact t-test and the genes past filtering from differential expression analysis were used as the background size.

96 97

SeRP in FMRP-FLAG neural progenitor cells We propose to circumvent anti-FMRP cross-reactivity to FXR2P by using a

FMRP-FLAG human progenitor stem cell line (hPSC) (Li, Shin et al. 2020). Using

CRISPR-Cas9, a 3XFLAG tag was integrated into the C-terminus of FMRP in a male hPSC line. First, we determined if FMRP-FLAG interacts with polysomes in a similar manner as wild-type FMRP. We prepared lysate from the FMRP-FLAG and parental hPSC lines and applied the lysate to a linear sucrose gradient for polysome analysis.

Figure III.9 shows that the polysome profile is similar between FMRP-FLAG and wild-type FMRP, indicating a lack of global changes in translation in the FMRP-FLAG hPSCs. We then extracted protein from each sucrose gradient fraction for western blot analysis and probed for RPS6, FMRP or FLAG. The association of wild-type

FMRP and FMRP-FLAG with polyribosomes are also similar, and the amount of RPS6 is also similar between the two cell lines (Figure III.9, bottom). Taken together, the polysome gradient analysis indicates that FMRP-FLAG fusion protein behaves similar to wild-type FMRP. The FMRP selective ribosome profiling experiment will be repeated in hPSCs with FMRP-FLAG. A FLAG antibody will be used to enrich for

FMRP-FLAG associated ribosomes, and the wild-type parental cells will serve as the negative control instead of IgG. This experimental design will allow us to identify mRNAs enriched by FMRP-associated ribosomes independent of the paralogues.

We observed enrichment of specific mRNAs by FMRP-associated ribosomes in wild-type mice and mice lacking FMRP, which may be explained by FMRP antibody cross-reactivity to FXR2P. Another possibility is that FMRP-associated

97 98

ribosomes recognize highly localized and unique binding to CDS that is missed by our whole gene analysis. Because we cannot perform reliable peak detection with only one FMRP SeRP library from Fmr1-/y mice, we will perform this analysis in the

FMRP-FLAG libraries which will have replicates.

Figure III.9: FMRP-FLAG cosediments with polysomes similar to wild-type FMRP

Polysome gradient analysis of wild-type FMRP (left) and FMRP-FLAG (right) hESCs with continuous absorbance at 254nm. Sedimentation is left to right in a continuous 10-50% (w/v) sucrose gradient. Global translation status is similar between wild- type and FMRP-FLAG hESCs. Bottom: western blots are shown for RPS6, FMRP, and FLAG from all fractions of the gradient. FMRP-FLAG cosediments with polyribosomes similar to wild-type FMRP.

98 99

Technical challenges to overcome for FMRP SeRP A common challenge in selective ribosome profiling experiments is minimizing post-lysis interactions and stabilizing the desired ribosome-complex of interest. When we run a sucrose gradient to observe FMRP co-sedimentation after

RNase digestion, we observe that most of the FMRP-polyribosome interaction depends on RNA, consistent with others publications (Figure III.10, top left and right) (Stefani, Fraser et al. 2004, Rachid El Fatimy 2016). Polysome gradient analysis reveals that a substantial population of FMRP moves to the free fraction after RNase digestion of mRNA, and there is no enrichment of FMRP in the monosome fraction, where the ~30nt RPFs are expected to cosediment. This result suggests that either a small percentage of the FMRP population is directly associated with ribosomes or that the FMRP-ribosome interaction is transient and/or stabilized by RNA. To address the latter concern, in vivo cross-linkers are frequently used in selective ribosome profiling experiments to stabilize ribosome-complex interactions. We treated HEK cells briefly with dithiobis succinimidyl propionate

(DPS), a cell-membrane permeable lysine-lysine cross-linker. In vivo cross-linking with DSP reveals mild stabilization of the FMRP-ribosome interaction; after RNaseI digestion, slightly more FMRP cosediments with the monosome fraction in DSP cross-linked samples compared to non-cross-linked samples when compared to the amount of monosome and RPS6 in the fractions (Figure III.10). However, even with

DSP crosslinking, most of FMRP still moves to the free fraction after RNase

99 100

digestion. Thus, an additional agent to stabilize protein-RNA interactions may be required, such as formaldehyde.

100 101

Figure III.10: Crosslinking stabilizes some FMRP-ribosome interaction

Polysome gradient analysis of HEK cells with continuous absorbance at 254nm. Sedimentation is left to right in a 10-50% (w/v) sucrose gradient. Top panel: noncross-linked HEK cells (-XL) with and without RNaseI digestion. Bottom panel: HEK cells treated with the in vivo crosslinking agent DSP (+XL) with and without RNase I digestion. Western blots of fractions from the entire gradient are shown for FMRP and RPS6. RNase I treatment collapses polysomes into monosomes, indicated with black square. DSP stabilizes some FMRP-ribosome interaction. XL = cross-linked.

101 102

Discussion

Overlapping function of FXRs and FMRP FMRP, FXR1P, and FXR2P make up the Fragile-X related family of proteins in mammals. FMR1 is the only member of the family that when inactivated, causes a human disease; as a consequence, FXR1/2 are relatively poorly studied. Because the brain targets of FXR1/2P are unknown, the targets specific to FMRP are also unknown. The FMRP family of related proteins is well conserved through the RNA- binding KH domains and 60S interaction region (Figure III.7), and thus the mRNA targets that depend on these regions should also largely overlap. In HEK cells, targets of ectopically expressed FXR1P and FXR2P overlap 95% with FMRP targets and are enriched for the same RNA recognition elements (Ascano, Mukherjee et al.

2012). Our FMRP selective ribosome profiling results also suggests that the targets enriched by FXR-associated ribosomes are the same, though to validate this we would need to perform FXR2P selective ribosome profiling.

Work from mice lacking Fxr2 support some overlapping functions with Fmr1; a combined Fxr2/Fmr1-KO mouse has a more severe phenotype than each individual KO mouse alone (Bontekoe, McIlwain et al. 2002, Spencer, Serysheva et al. 2006, Saré, Figueroa et al. 2019). In addition, synaptic localization of FXR2P is disrupted in Fmr1 knockout mice (Wang, Smith et al. 2016) but not FXR1P, indicating that FXR2P contributes to the synaptic dysfunction of FXS mice. Fxr2-KO mouse models also suggests that FXR2P is required for proper brain development

102 103

and function, as FXR2P deficiency impairs neurogenesis in the dentate gyrus of the hippocampus by reducing Noggin mRNA stability (Guo, Zhang et al. 2011).

Why are mutations in FMR1 associated with a human disease and not FXR1/2 if the FXR family is well conserved? FXS is caused by an unstable GCC trinucleotide repeat expansion in the 5′UTR of FMR1 and this repeat is not present in FXR1 or

FXR2, indicative of difference in disease mechanism (Verkerk, Pieretti et al. 1991).

Also, although FXR1P and FXR2P have some overlapping function with FMRP based on work in mouse models, the paralogues do not completely compensate for loss of

FMRP (Spencer, Serysheva et al. 2006, Saré, Figueroa et al. 2019). This indicates a unique function of FMRP. The RGG box and C-terminus of FMRP are more divergent from the paralogues (Figure III.7), leading to the speculation that these regions are important for FMRP-specific functions over the paralogues. One possibility could be that the FXRs bind similar mRNAs but have different functional outcomes and regulation. Additional unbiased transciptome methods such as CLIP of FXR1P/2P and RNA-seq of mouse models are needed to distinguish the unique and overlapping functions of the paralogues from FMRP.

Auto-regulation of Fmr1 translation by FMRP Our FMRP selective ribosome profiling reveals that FMRP enriches for RPFs from its own mRNA, suggesting that FMRP regulates it’s own mRNA through the ribosome. FMRP binds with high affinity to its own mRNA, which is mediated by

FMRP directly binding to the RGG-coding region of FMR1 to repress protein synthesis (Ashley, Wilkinson et al. 1993, Schaeffer, Bardoni et al. 2001). Our FMRP

103 104

selective ribosome profiling indicates that the peak enrichment of Fmr1 RPFs occurs before the RGG-coding region, suggestive of distinct RNA-dependent and ribosome- dependent functions of FMRP in regulating Fmr1. Collectively, these data supports the model proposed by Chen et al of how FMRP may simultaneously bind mRNA and the ribosome: FMRP binds to G-quadruplex containing mRNAs, such as the RGG region of Fmr1, and then docks on the ribosome to repress translation (Chen,

Sharma et al. 2014). The outcome of FMRP binding to ribosomes on its on mRNA may be translational repression and/or mRNA decay.

FMRP-ribosome interaction We initiated this project based on the model that FMRP may be directly associated with the ribosome as proposed by Chen et al (Chen, Sharma et al. 2014).

Our initial results supported this model based on the observation that FMRP sediments with the ribosomal protein pellet and FMRP co-IPs with ribosomal proteins after RNAse digestion. Sucrose gradient analysis, which is higher resolution than a sucrose cushion, reveals that most of the FMRP-polysome interaction is dependent on RNA, which is consistent with data from others (Stefani, Fraser et al.

2004, Rachid El Fatimy 2016). Differences in the sucrose cushion and gradient results can arise from the different populations of FMRP captured by both methods.

A sucrose cushion will capture all proteins, whereas a sucrose gradient misses the heavier sedimenting ribosomes of long mRNAs at the bottom of the tube, where

FMRP is enriched in. Regardless, our results suggest that the RPFs captured by our

104 105

FMRP-IP selective ribosome profiling represent a small population of direct FMRP- ribosome complexes.

Our results suggest that that a small population of FMRP may bind directly to the ribosome and/or the FMRP-ribosome interaction may be transient and stabilized by RNA (rRNA and/or mRNA). The latter interpretation is at odds with the reported high affinity (KD ~ 20nM) that dFMRP binds to ribosomes from Chen et al (Chen, Sharma et al. 2014). Because multiple studies find that FMRP can interact directly with select ribosomal proteins and ribosomal RNA, one possibility is that

FMRP may directly bind to ribosomes via rRNA, and RNase digestion disrupts this interaction (Ishizuka, Siomi et al. 2002). However, it is unclear if FMRP binding to rRNA occurs in a specific manner. Based on the model by Chen et al that FMRP binds the intersubunit space of the ribosome, FMRP may be in close enough proximity to cross-link to tRNAs. Photoactivatable Ribonucleoside-Enhanced Crosslinking and

Immunoprecipitation (PAR-CLIP) is a method that reveals high resolution RBP binding based on T to C conversions in cDNA (Hafner, Landthaler et al. 2010).

Reanalysis of FMRP PAR-CLIP datasets may reveal if FMRP binds rRNA and tRNA in a specific manner and provide support if FMRP directly interacts with ribosomes

(Ascano, Mukherjee et al. 2012). For example, the RBP Zinc Finger Protein 598

(ZNF598) is a ribosome-associated protein that participates in ribosome-associated quality control, a pathway to disassemble ribosomes after translation of poly(A) residues. ZNF598 shares a similar binding pattern as FMRP, where most CLIP sites occur in the CDS (Garzia, Jafarnejad et al. 2017). PAR-CLIP experiments reveals that

105 106

ZNF598 binds to specific sites on 5S rRNA and is enriched for binding to tRNALys(UUU), consistent with a role of ZNF598 in sensing AAA codons on ribosomes.

106 107

Materials and methods

Sucrose cushion

Roughly 9 million HEK cells were lysed in 1mL of polysome lysis buffer (20mM Tris-

HCl pH 7.4, 100mM KCl, 5mM MgCl2 100ug/mL cycloheximide, 1mM DTT, 1X EDTA- free protease inhibitory, and 0.3% Triton-X100) in a Dounce homogenizer. The

A260 was measured, and the lysate was treated with 1ng, 10ng, or 100ng RNase

A/A260 for 30min at 25C with 300rpm rotation. The lysate was then layered onto a sucrose cushion (0.5M and 1M of sucrose dissolved in polysome lysis buffer without

Triton-X100) and centrifuged at 80,000 for 3hr at 4C in a TL1-100.3. The pellet was dissolved in equal volume as the supernatant fraction in a TBS solution with 1% SDS pre-warmed to 95C. The pellet was incubated for 5min and mixed to ensure resuspension of the pellet.

In vivo DSP cross linking

On the day before cell lysis, HEK cells were split into a 150mm dish to reach 80% confluency on the day of experiment. The media was aspirated and the cells were washed 2X with 15mL of warmed PBS containing 100ug/mL of cycloheximide. The media was removed thoroughly to avoid crosslinking of free amino acids in the media. Cells were crosslinked with 15mL of 1mM DSP dissolved in PBS at room temperature for 1min with gentle rocking. The DSP solution was removed, and the crosslinking reaction was quenched with 15mL of 50mM Tris, pH 7.4 at room temperature for 1min with gentle rocking. The Tris solution was removed, and the

107 108

cells were washed 2X with 15mL of ice-cold PBS containing 100ug/mL of cycloheximide. Cells were lysed directly with 500uL of ice-cold polysome buffer

(20mM Tris-HCl pH 7.5, 5mM MgCl2, 100mM KCl, 1mM DTT, 100ug/mL CHX, 1X

EDTA-free protease inhibitor, 1% NP-40) added to the plate. Cells were scraped from the plate and transferred to a 1.5mL tube. The cells were triturated through a

25G needle 10X in the cold room and incubated on ice for 10min. The lysate was spun at 14,000g for 10min at 4C. The supernatant was collected and the RNA concentration measured by Nanodrop and Qubit RNA Broad Range assay. For the

RnaseI digested samples, RNaseI was added at 100U/120ug of RNA for 5min at 4C with end-over-end rotation. SuperRnaseIn (2.5uL) was added to stop the RNase digestion. The lysate was layered onto a sucrose gradient was described below.

hPSC cell lysis and polysome profiling analysis

Frozen hPSCs cell pellets were thawed in polysome buffer (20mM Tris-HCl, pH 7.4,

5mM MgCl2, 100mM KCl, 1mM DTT, 100ug/mL cycloheximide, 1X EDTA-free protease inhibitor (Sigma, #11836170001), and 1% TritonX-100 w/v) for 5min on ice and the resuspended. A 25G needle was used to triturate the cells 10X and incubated on ice for 10min. The homogenate was centrifuged at 14,000g for 10min at 4°C. The supernatant was collected and the lysate was quantified using a

Nanodrop. The samples were adjusted to equal concentrations of RNA using polysome buffer and equal amounts of lysate was layered over a 10-50% w/v sucrose gradient prepared in polysome buffer with 100ug/mL cycloheximide and

108 109

centrifuged at 35K rpm for 2.5hr at 4°C using a SW41 rotor. The sucrose gradients were then fractionated from top to bottom and monitored with continuous UV absorbance (A254). For downstream western blots, equal amounts were taken from each fraction.

Western blot analysis

4X Laemelli buffer (8% sodium dodecyl sulfate, 20% glycerol, 0.008% bromophenol blue, 250mM Tris-HCl pH 6.8, 800mM dithiothreitol) was added to samples to reach a 1X concentration and heated at 95°C for 10min. Samples were loaded onto a 10%

SDS-PAGE gel and ran at 80V until samples reached the end of the stacking gel, which was then run at 120V. The proteins were transferred to a PVDF membrane for 2 hours at 100V in the cold room. The membranes were blocked in 5% milk in

TBS-T for 1 hour at room temperature and incubated with primary antibody in blocking buffer overnight at 4°C. The blots were washed 4X for 5 minutes each with

TBS-T and incubated with HPR conjugated secondary antibody in 5% milk in TBS-T for 1 hour at room temperature. For immunobloting of IP experiments, TrueBlot rabbit secondary HPR-conjugated antibody was used (Rockland, #18-8816-33).

Blots were washed 4X for 5 minutes each before development with ECL. The following primary antibodies were used for the study: FMRP (Abcam, #ab17722),

FXR2 (BD Biosciences, #611330), RPS6 (Cell Signaling, #2217S), RPL4 (Proteintech,

#11302-1-AP), FLAG M2 antibody (Sigma, #F1804).

109 110

RPL4 and FMRP immunoprecipitation

Male wild-type C57BL/6 mice and Fmr1tmCgr mice (Jackson Laboratory, stock number 003025) were used (P22-23). Brains were rapidly removed and quickly washed in ice-cold 1X Hank’s Buffered Saline Solution with 100ug/mL cycloheximide. The mid-brain, corpus callosum, and hippocampi were dissected from the cortices. Two hemispheres of the cortices were added to a Dounce homogenizer with 1mL of polysome lysis buffer (20mM Tris-HCl, 100mM KCl,

10mM MgCl2, 1mM DTT, 100ug/mL CHX, and 1X EDTA-free protease inhibitor). The cortices were lysed with 8 strokes each of A (loose) and B (tight) pestles. The lysate was transferred to a 1.5mL tube and NP-40 added to a final 1% v/v. The lysate was incubated on ice for 10min, centrifuged at 2,000g for 10min at 4C. The supernatant was collected and spun at 16,000g for 10min at 4C. The supernatant was collected and the A260 measured on the Nanodrop. For the wild-type and Fmr1-KO samples, the lysate concentration was adjusted to be equal. The lysate was digested with

100ng /A260 of RNase A for 30min at 25C with end-over-end rotation. After RNase reaction, SuperRNasIn was added. The lysate was centrifuged at 10,000g for 5min at

4C, and the supernatant was collected for IP. Roughly 1.3mL of lysate was precleared for 30min at 4C with washed Protein A Dynabeads resuspended in lysis buffer. The lysate was then equally split for the RPL4, FMRP, or IgG IP, and 5ug of antibody was added directly to the lysate, corresponding to roughly 5 A260 units per IP. The antibody and lysate was incubated for 3hr at 4C with end-over-end rotation. 50uL of washed Protein A Dynabeads were then added to the lysate and

110 111

incubated for 1hr at 4C with end-over-end rotation. The unbound supernatant was removed, and the beads were washed with 1mL of cold polysome buffer for 5min with end-over-end rotation for a total of 4 times. The last wash was transferred to a new tube to avoid eluting nonspecifically bound proteins and mRNAs. RNA was eluted with 1mL of Trizol and protein was eluted with 1X Laemmeli buffer (2% sodium dodecyl sulfate, 10% glycerol, 0.002% bromophenol blue, 62.5mM Tris-HCl pH 6.8, 200mM dithiothreitol) and heated for 10 minutes at 95C.

Ribosome profiling library prep and sequencing analysis

RNA was extracted following the Trizol protocol and ribosome profiling libraries were generated as described in Chapter 2. Read filtering and mapping was described in Chapter 2. DESeq2 was used to calculate differential expression analysis. Because there were no differences in the FMRP-IP between wild-type and

Fmr1-/y samples, these libraries were combined as FMRP-IP samples. FMRP-IP libraries were compared to RPL4-IP libraries, and both FMRP-IP and RPL4-IP mRNAs had to be enriched 4-fold over IgG negative control to be considered enriched. Transcript plots were generated using reads normalized by counts per million. Examples of transcript plots were for the most abundant isoform. Publically available datasets used in this chapter are from the following publications: (Darnell,

Van Driesche et al. 2011, Ascano, Mukherjee et al. 2012)

111 112

Acknowledgements

We thank Botao Liu and Emily Stackpole for technical help and advice. We thank

Meng Li and Xinyu Zhao for providing the FMRP-FLAG cells.

112 113

Table III-2: List of 125 mRNAs enriched in anti-FMRP ribosomes

Symbol Gene Name baseMean log2FC padj

1810055G

02Rik RIKEN cDNA 1810055G02 gene 576.19 1.68 2.79E-56

Aff2 AF4/FMR2 family, member 2 1034.63 0.81 2.57E-12

Aff4 AF4/FMR2 family, member 4 4494.59 0.76 9.36E-11

angiogenic factor with G patch and FHA

Aggf1 domains 1 463.50 1.05 4.37E-16

Ankle2 ankyrin repeat and LEM domain containing 2 1399.93 0.91 3.61E-22

Ankrd12 ankyrin repeat domain 12 2461.85 1.33 2.06E-15

Ankrd26 ankyrin repeat domain 26 476.64 0.86 3.06E-07

amyloid beta (A4) precursor protein-binding,

Apba2 family A, member 2 11535.38 1.73 1.56E-51

Apc adenomatosis polyposis coli 5861.91 0.98 4.99E-14

Arhgap24 Rho GTPase activating protein 24 123.39 0.96 8.85E-08

Asxl3 additional sex combs like 3 (Drosophila) 224.74 1.16 2.11E-13

Atxn1l ataxin 1-like 804.53 2.51 2.95E-97

Bahcc1 BAH domain and coiled-coil containing 1 302.33 1.23 4.23E-16

Bzrap1 benzodiazepine receptor associated protein 1 730.54 0.82 9.16E-14

Cacna1a calcium channel, voltage-dependent, P/Q type, 3652.75 0.86 1.60E-10

113 114

alpha 1A subunit

Cage1 cancer antigen 1 90.41 1.09 7.55E-08

core-binding factor, runt domain, alpha subunit

Cbfa2t3 2, translocated to, 3 (human) 312.90 0.77 2.53E-08

Ccdc61 coiled-coil domain containing 61 165.67 1.04 3.08E-11

Cep162 centrosomal protein 162 225.17 0.76 1.69E-05

Cep170b centrosomal protein 170B 4892.02 0.78 4.42E-12

Cep85l centrosomal protein 85-like 407.50 0.80 2.61E-08

Chgb chromogranin B 12245.34 0.90 1.10E-06

Cic capicua transcriptional repressor 3099.18 0.96 3.67E-13

Cobl cordon-bleu WH2 repeat 8003.88 2.15 1.73E-75

cytoplasmic polyadenylation element binding

Cpeb4 protein 4 2578.38 1.00 3.31E-26

cysteine and glycine-rich protein 2 binding

Csrp2bp protein 457.58 1.33 5.52E-30

Dbn1 drebrin 1 2881.80 1.51 7.64E-18

Diap2 NA 1086.83 0.89 1.56E-18

Dis3l DIS3 like exosome 3'-5' exoribonuclease 754.57 1.37 2.51E-24

Dlg5 discs, large homolog 5 (Drosophila) 302.32 0.77 1.94E-06

Dnajc21 DnaJ heat shock protein family (Hsp40) 241.09 0.94 3.39E-09

114 115

member C21

Donson downstream neighbor of SON 265.40 0.99 2.15E-10

Dzip1 DAZ interacting protein 1 106.78 0.77 4.82E-05

E130308A

19Rik RIKEN cDNA E130308A19 gene 368.56 1.90 1.55E-36

Ears2 glutamyl-tRNA synthetase 2, mitochondrial 378.14 1.77 2.03E-32

Ep300 E1A binding protein p300 2366.79 1.12 2.91E-31

Etv1 ets variant 1 2921.93 1.00 1.30E-23

Fam178a family with sequence similarity 178, member A 372.39 0.97 6.69E-11

Fbxo34 F-box protein 34 1181.79 0.90 5.76E-07

Foxp1 forkhead box P1 3828.83 1.23 1.59E-32

fragile X mental retardation, autosomal

Fxr2 homolog 2 2993.93 1.56 1.52E-21

Gigyf1 GRB10 interacting GYF protein 1 2695.72 1.32 2.84E-09

Gigyf2 GRB10 interacting GYF protein 2 2732.45 1.76 3.24E-51

G-protein signalling modulator 1 (AGS3-like, C.

Gpsm1 elegans) 2707.45 1.77 1.58E-31

Gsn gelsolin 6627.25 0.79 8.28E-07

Heatr3 HEAT repeat containing 3 1384.23 1.04 3.79E-24

Hgf hepatocyte growth factor 563.28 1.53 7.26E-16

115 116

Ift172 intraflagellar transport 172 1430.96 0.77 1.79E-14

Il16 interleukin 16 114.38 0.98 7.11E-08

Irf9 interferon regulatory factor 9 121.98 1.02 4.29E-09

Jade3 jade family PHD finger 3 262.10 1.69 1.64E-35

Kdm4a lysine (K)-specific demethylase 4A 2304.25 1.27 1.88E-37

Kdm5c lysine (K)-specific demethylase 5C 1700.45 0.99 1.66E-24

Kif1b kinesin family member 1B 18793.61 0.75 2.97E-19

Kif21b kinesin family member 21B 1076.81 0.77 5.11E-12

Klf15 Kruppel-like factor 15 694.06 1.15 4.28E-21

Leo1, Paf1/RNA polymerase II complex

Leo1 component 1251.26 1.01 4.14E-17

Lrrc7 leucine rich repeat containing 7 3586.39 0.97 2.15E-36

mitogen-activated protein kinase kinase kinase

Map4k3 kinase 3 1661.95 1.02 2.52E-21

mitogen-activated protein kinase 8 interacting

Mapk8ip3 protein 3 8072.36 0.91 4.80E-16

Mark2 MAP/microtubule affinity regulating kinase 2 3073.59 1.02 1.21E-20

microtubule associated serine/threonine kinase

Mast4 family member 4 1547.78 1.50 1.76E-41

Mepce methylphosphate capping enzyme 878.18 0.81 1.57E-13

116 117

Mgat5 mannoside acetylglucosaminyltransferase 5 2005.33 0.79 1.41E-17

Myh10 myosin, heavy polypeptide 10, non-muscle 9154.52 1.00 4.14E-09

Myh14 myosin, heavy polypeptide 14 481.89 0.82 3.21E-08

Myh9 myosin, heavy polypeptide 9, non-muscle 2311.50 1.10 2.44E-12

Myo5a myosin VA 13206.64 0.82 6.62E-07

Myo9b myosin IXb 306.78 0.78 2.69E-06

Ncoa2 nuclear receptor coactivator 2 3784.40 1.62 2.36E-77

Ncor1 nuclear receptor co-repressor 1 3223.13 1.08 2.41E-13

Nefh neurofilament, heavy polypeptide 3356.17 1.88 2.22E-39

Nipbl Nipped-B homolog (Drosophila) 1735.90 1.59 1.97E-59

Notch1 notch 1 957.30 1.02 1.16E-08

Optn optineurin 919.50 1.76 1.83E-45

Pard3 par-3 family cell polarity regulator 462.95 1.38 7.05E-26

PCF11 cleavage and polyadenylation factor

Pcf11 subunit 764.77 0.79 2.58E-08

Pcm1 pericentriolar material 1 4243.03 0.78 1.69E-12

Phactr2 phosphatase and actin regulator 2 424.24 1.76 1.45E-34

polymerase (RNA) mitochondrial (DNA 1.30E-

Polrmt directed) 1489.95 2.40 102

Ppargc1a peroxisome proliferative activated receptor, 1348.91 1.61 2.19E-48

117 118

gamma, coactivator 1 alpha

Ppip5k2 diphosphoinositol pentakisphosphate kinase 2 476.55 1.70 9.13E-47

protein phosphatase 1, regulatory (inhibitor)

Ppp1r9a subunit 9A 1128.85 1.06 5.72E-14

protein kinase, DNA activated, catalytic

Prkdc polypeptide 1968.06 1.44 1.57E-15

protein tyrosine phosphatase, non-receptor

Ptpn2 type 2 577.18 1.18 3.43E-15

Rasgrp2 RAS, guanyl releasing protein 2 5395.77 1.76 1.14E-16

RNA binding protein, fox-1 homolog (C.

Rbfox2 elegans) 2 6220.95 1.68 1.17E-66

Rbm12b1 RNA binding motif protein 12 B1 91.55 1.07 4.38E-08

Rbm12b2 RNA binding motif protein 12 B2 158.43 0.83 4.43E-06

Rexo1 REX1, RNA exonuclease 1 658.14 1.02 2.00E-08

Rgag4 retrotransposon gag domain containing 4 632.29 1.11 1.35E-25

Rnf115 ring finger protein 115 1989.70 0.82 3.76E-16

Rnf185 ring finger protein 185 824.41 0.96 9.67E-21

runt-related transcription factor 1; translocated

Runx1t1 to, 1 (cyclin D-related) 904.65 1.00 2.58E-20

Sap130 Sin3A associated protein 4473.83 2.33 5.99E-50

118 119

Setd4 SET domain containing 4 95.62 0.96 4.98E-05

Sgpp1 sphingosine-1-phosphate phosphatase 1 765.28 0.94 9.63E-18

Sipa1l2 signal-induced proliferation-associated 1 like 2 933.91 0.75 2.87E-14

Sptbn1 spectrin beta, non-erythrocytic 1 4553.37 0.86 2.14E-06

Srcin1 SRC kinase signaling inhibitor 1 9607.66 1.30 3.36E-30

Tbc1d14 TBC1 domain family, member 14 925.71 1.00 6.65E-25

Tet2 tet methylcytosine dioxygenase 2 189.27 0.80 1.50E-06

Thbs1 thrombospondin 1 154.52 0.98 2.62E-07

Thbs2 thrombospondin 2 390.19 1.11 8.54E-09

0.00223

Thrap3 thyroid hormone receptor associated protein 3 5146.33 0.75 9933

Tjap1 tight junction associated protein 1 142.86 1.11 7.62E-11

Tnip1 TNFAIP3 interacting protein 1 1392.44 2.09 6.55E-76

Tnrc18 trinucleotide repeat containing 18 1613.71 1.17 6.46E-19

Tnrc6b trinucleotide repeat containing 6b 2374.83 1.20 2.59E-11

Triobp TRIO and F-actin binding protein 222.01 0.94 2.40E-05

transformation related protein 53 binding

Trp53bp2 protein 2 884.02 0.83 1.47E-09

Ubap2l ubiquitin-associated protein 2-like 4544.19 1.02 2.12E-10

Zbtb38 zinc finger and BTB domain containing 38 1932.08 1.27 9.44E-45

119 120

Zdbf2 zinc finger, DBF-type containing 2 536.07 0.98 5.82E-10

Zfp106 zinc finger protein 106 1662.22 1.31 7.99E-43

Zfp14 zinc finger protein 14 141.03 0.79 3.54E-06

Zfp191 NA 1088.38 0.79 4.11E-12

Zfp40 zinc finger protein 40 263.96 0.75 7.01E-07

Zfp407 zinc finger protein 407 626.89 0.80 3.33E-08

Zfp52 zinc finger protein 52 81.59 1.09 1.02E-07

Zfp800 zinc finger protein 800 535.66 0.80 4.43E-12

Zfyve9 zinc finger, FYVE domain containing 9 1257.06 0.76 2.50E-12

Zkscan3 zinc finger with KRAB and SCAN domains 3 425.58 0.97 2.09E-10

1.63E-

Znfx1 zinc finger, NFX1-type containing 1 5528.45 2.80 185

Zscan12 zinc finger and SCAN domain containing 12 806.98 3.11 2.12E-93

120 121

CHAPTER IV DISCUSSION AND CONCLUSIONS

Summary

mGluR5 signaling is required for a form of synaptic plasticity (mGluR-LTD) that is dysregulated in multiple mouse models of autism, including TSC and FXS. Previous work found that Tsc2+/- mice, which model TSC autism, have decreased protein synthesis in the hippocampus and deficient LTD. Because restoring protein synthesis corrects the decreased LTD in Tsc2+/- mice, identifying the mRNAs dysregulated in these animals may provide insight into their impaired synaptic function. In addition, the mRNAs responsive to mGluR5 signaling are largely unknown, and their identities are important because they may underlie the altered mGluR-LTD in multiple mouse models of autism.

Our project sought to identify mRNAs responsive to mGluR5 signaling in the normal and

Tsc2+/- hippocampus by using unbiased, high resolution approaches: RNA-seq and ribosome profiling. We find that targets of the RNA-binding protein FMRP have increased RNA abundance in Tsc2+/- mice as well as following mGluR5 signaling.

Conversely, in Fmr1-/y mice, a previous study finds that FMRP binding targets have decreased mRNA levels (Shah, Molinaro et al. 2020). mGluR-LTD activates eEF2K/eEF2 signaling to converge on slowing ribosome movement, and we find that

Tsc2+/- mice have impaired mGluR5-eEF2K signaling. Pharmacologically slowing ribosomes with cycloheximide rescues deficient mGluR-LTD in Tsc2+/- mice, suggesting a novel link between ribosome translocation and synaptic plasticity deficits in these animals.

121 122

FMRP binding targets connects mGluR-LTD with two mouse models of autism

A goal of our project was to identify the mRNAs dysregulated in TSC autism mice and to determine if these mRNAs overlap with other mouse models of autism. Our results suggest that FMRP binding targets are altered bidirectionally in Tsc2+/- and Fmr1-

/y mice and regulated by mGluR-LTD. The group of FMRP binding targets may explain the findings from a previous study, where a combined Tsc2+/- and Fmr1-/y mouse model can rescue protein synthesis, mGluR-LTD and behavior to wild-type levels (Auerbach,

Osterweil et al. 2011). To assess whether this group of mRNAs are responsible for the phenotypes of the TSC and FXS mice, one should determine whether FMRP binding targets return to wild-type levels in a combined Tsc2+/-and Fmr1-/y mouse. Interestingly, there is a correlation between FMRP binding stringency and the degree of change in

Tsc2+/- animals and following mGluR-LTD, where mRNAs bound by more FMRP increase in abundance more than mRNAs bound with less FMRP. This would suggest that TSC2 regulates FMRP activity, perhaps through altered signaling. We do not observe changes in mRNA, RPF, or protein levels of FMRP in Tsc2+/- mice, suggesting that the increased abundance of FMRP binding targets could arise from changes in FMRP activity, for example through phosphorylation. Following mGluR5 signaling, dephosphorylation of FMRP promotes its release from miRNA-RISC complexes and relieves translational repression on its target PSD-95 mRNA, a synaptic scaffold

(Nalavadi, Muddashetty et al. 2012). Phosphorylation of FMRP can also enhances phase

122 123

separation with RNA into granules to promote deadenylation and translational repression

(Kim, Tsang et al. 2019, Tsang, Arsenault et al. 2019).

TSC2: potential regulator of mGluR-LTD signaling to eEF2 and FMRP

TSC2 is an inhibitor of mTORC1 signaling, so our initial hypothesis was that increased mTORC1 signaling to 4E-BP1 and S6K1 dependent translation would mediate the changes in Tsc2+/- animals. However, we do not find clear evidence of such translational dysregulation (i.e., no enrichment of TOP motifs or long, structured 5′UTR in differentially expressed mRNAs). Although we observe a trend for increased phospho- rpS6 in Tsc2+/- mice, a common readout for mTORC1-S6K1 activation, the role of phospho-rpS6 on translation is unclear. Instead, our work finds that signaling to eEF2K/eEF2 phosphorylation, a signaling pathway required for mGluR-LTD, is basally elevated in Tsc2+/- mice and no longer responsive to mGluR5 signaling. Although eEF2K is a substrate of mTORC1-S6K1 signaling, the elevated levels of phospho-eEF2 is opposite of the direction predicted by mTORC1 activation. This suggests that the dysregulated eEF2K/eEF2 signaling in Tsc2+/- mice may be from an mTORC1- independent process. Increased levels of phospho-eEF2 inhibit general translation, and our results are consistent with a previous finding that Tsc2+/- mice have decreased protein synthesis instead of the predicted elevated translation from mTORC1 hyperactivation

(Auerbach, Osterweil et al. 2011).

The altered eEF2K/eEF2 signaling in Tsc2+/ animals may link the changes in

FMRP binding targets and impaired synaptic plasticity in these mice. mGluR5-eEF2K and mGluR5-FMRP signaling regulates synthesis of specific proteins required for 123 124

mGluR-LTD, but previous work suggest that these pathways are parallel and distinct

(Park, Park et al. 2008). Because we observe elevated levels of FMRP binding targets that correlates with increased eEF2 phosphorylation, our results suggest that TSC2 may be upstream of mGluR5-eEF2K and mGluR5-FMRP signaling. Thus, TSC2 may link these two important signaling pathways required for mGluR-LTD.

Altered eEF2K/eEF2 signaling may underlie other phenotypes in Tsc2+/- mice

Our results demonstrate a novel link between dysregulated eEF2K/eEF2 signaling and a mouse model of TSC autism, where phospho-eEF2 levels are basally elevated and no longer responsive to mGluR5 activation. A future direction of inquiry would be to determine if targeting eEF2K/eEF2 signaling could ameliorate other phenotypes of the

Tsc2+/- mice. For example, eEF2K/eEF2 is required for multiple forms of protein synthesis-dependent synaptic plasticity, learning and memory, and maintaining the excitatory/inhibitory balance in the brain, all of which show evidence of disruptions in

Tsc2+/- animals (Sutton, Taylor et al. 2007, Park, Park et al. 2008, Gildish, Manor et al.

2012, McCamphill, Farah et al. 2015, Weng, Chen et al. 2016, Heise, Taha et al. 2017,

Antoine, Langberg et al. 2019). Our data suggest that slowing ribosome movement with cycloheximide to mimic phospho-eEF2 can rescue mGluR-LTD in Tsc2+/- mice.

However, a limitation to this interpretation is that cycloheximide has effects independent of inhibiting ribosome movement, such as internalization of surface epidermal growth factor receptors and inhibiting actin cytoskeletal rearrangement (Oksvold, Pedersen et al.

2012, Chang, Tsai et al. 2013, Darvishi and Woldemichael 2016). Additional experiments

124 125

to target eEF2K specifically would further support if restoring eEF2K/eEF2 signaling rescues mGluR-LTD, behavior, and protein synthesis in Tsc2+/- mice.

Crosstalk of eEF2K/eEF2 signaling and TOP mRNAs in Tsc2+/- animals

The TOP mRNAs are a distinct group that contain an oligopyrimidine tract in the

5′UTR, encode many components of the translation machinery, and are sensitive to mTORC1 regulation (Thoreen, Chantranupong et al. 2012). In Tsc2+/- animals, we find that the TOP mRNAs have decreased mRNA and mildly increased RPFs levels. Although the slight increase in translation of TOP mRNAs would be consistent with increased mTORC1 signaling, the decrease in RNA abundance is more dramatic and overall, likely leads to decreased levels of ribosomal proteins. Thus, it is unclear if initial mTORC1 hyperactivation drives excessive TOP mRNA translation in Tsc2+/- animals, which then compensates by decreasing RNA levels. Regardless, we observe evidence of dysregulated eEF2K/eEF2 signaling and the TOP mRNAs in Tsc2+/- animals, which may be mediated through the RBP LARP1. LARP1 directly binds TOP mRNAs to bidirectionally function as an activator or repressor of mRNA stability and translation based on phosphorylation status (Gentilella, Moron-Duran et al. 2017). There is evidence of cross talk between eEF2 phosphorylation, LARP1, and regulation of TOP mRNAs, where deficiency in ribosomal proteins activates a ribosomal stress response that increases eEF2 phosphorylation (Gismondi, Caldarola et al. 2014). LARP1 deficiency can also decrease ribosome levels, activate a ribosomal stress response, and increase eEF2K/eEF2 signaling

(Gentilella, Moron-Duran et al. 2017). The role of LARP1 is unexplored in the brain, and a future direction of inquiry would be to determine if LARP1 activity is altered in Tsc2+/- 125 126

mice and mediates the decrease in TOP mRNA levels. Of note, we do not have evidence of altered mRNA or RPF levels of Larp1 in Tsc2+/- animals. An unbiased approach such as phosphoproteomics could identify changes in LARP1 phosphorylation and other components of TSC-Rheb signaling in the Tsc2+/- mice.

Role of mRNA turnover and stability in mGluR-LTD

mGluR-LTD is a protein-synthesis dependent form of synaptic plasticity that is dysregulated in multiple mouse models of autism and shows promise in ameliorating

ASD-like behaviors (Bear, Huber et al. 2004, Kelly, Schaeffer et al. 2018). However, the identities of the mGluR5-responsive mRNAs in the intact hippocampus are largely unknown. Using unbiased RNA-seq and ribosome profiling, our study identified these mGluR5-responsive mRNAs. At 5 minutes after induction of mGluR5 signaling, we observe a rapid increase in mRNAs that are enriched for synaptic function and FMRP binding targets. The current model in the field is that FMRP acts as a translational repressor of mRNAs required for mGluR-LTD. Our findings are consistent with a model of FMRP as a regulator of mGluR5-responsive mRNAs, but also suggests a role of

FMRP in regulating mRNA turnover. Under basal conditions, FMRP may promote degradation of specific mRNAs. Upon mGluR5 activation, other labs have shown that

FMRP is inactivated by post-translational modifications and degradation, which may allow for increased stabilization of mRNA levels (Nalavadi, Muddashetty et al. 2012,

Huang, Ikeuchi et al. 2015).

Traditionally, translational control of mGluR-LTD has been the main focus of research, and our observation of rapidly increased and decreased mRNAs following 126 127

mGluR5 signaling was surprising, because we expected a stronger translational response.

Because the transcriptional inhibitor actinomycin D does not block mGluR-LTD (arguing that RNA synthesis is not required) and because we did not observe induction of immediate early genes (Arc, c-fos, Egr-1), we speculate that the rapid changes in mRNA abundance is from altered mRNA turnover. An RNA metabolic labeling assay is required to demonstrate altered mRNA turnover following mGluR-LTD in hippocampal slices, although this is technically challenging to perform in tissue.

Our study suggests that regulators of mRNA turnover are likely to be important for mGluR5 signaling. One such example could be miRNAs. Synaptic stimulation can promote rapid degradation of miRNAs or dissociation of miRNAs from mRNA targets to prevent mRNA degradation and relieve translational repression. In retinal neurons, multiple miRNAs are quickly degraded following a light stimulus (Krol, Busskamp et al.

2010). It is unknown if rapid degradation of specific miRNAs occurs after mGluR5 signaling and if so, whether these miRNAs are required for mGluR-LTD. miRNA association with targets can also be rapidly controlled by synaptic activity.

Phosphorylated FMRP interacts with miR-125a-RISC to repress translation of the synaptic scaffold PSD-95 mRNA (Edbauer, Neilson et al. 2010, Muddashetty, Nalavadi et al. 2011). Upon mGluR5 signaling, FMRP is dephosphorylated and releases miR-125a-

RISC from PSD-95 mRNA (Muddashetty, Nalavadi et al. 2011). Because FMRP targets increase in abundance following mGluR5 signaling in our data, an unanswered question is whether miR-125a or other miRNAs are required for stabilization of these mRNAs following mGluR-LTD.

127 128

Another mechanism of rapid mRNA turnover is through control of the poly(A) tail. Shortening of the poly(A) tail often leads to destabilization of the mRNA by recruitment of decay factors. Lengthening of the poly(A) tail promotes circularization of the mRNA for initiation by facilitating the interaction of polyA Binding Protein (PABP) with eukaryotic initiation factor 4E (eIF4E), the cap-binding protein. Thus, lengthening of the poly(A) tail length can allow for both rapid stabilization and translation of mRNAs, and the importance of this process has been demonstrated for other forms of synaptic plasticity. The deadenylase CNOT7 is quickly degraded following chemical

LTP in cultured neurons, and this correlates with an increase of dendritic poly(A) signal within 30 seconds, which is likely from rapid mRNA stabilization and increases in poly(A) tail length (McFleder, Mansur et al. 2017). Thus, mGluR5 signaling may promote poly(A) tail lengthening by inactivating deadenylases, activating poly(A) polymerases, and/or dissociation of miRNAs to increase stability and translation of target mRNAs. One method to support if mRNAs are stabilized following mGluR5 activation is to measure poly(A) tail length of specific mRNAs by an unbiased method such as TAIL-

Seq (Chang, Lim et al. 2014). This approach is technically easier to perform than metabolic labeling of hippocampal slices to measure mRNA turnover. To determine whether mRNA stabilization is required for mGluR5 signaling, ideally one would acutely inhibit mRNA turnover. However, this approach is challenging because of the lack of chemical inhibitors that specifically block mRNA degradation. Remaining questions include if specific deadenylases and poly(A) polymerases are regulated by mGluR5

128 129

signaling and if so, do they interact with FMRP to control abundance of mRNAs required for mGluR-LTD.

Limitations of mouse models of TSC and whole tissue sequencing

A limitation of our study is identifying changes within a heterogeneous tissue such as the brain, particularly for Tsc2+/- mice where the disturbances are mild. We chose a germline heterozygote model to reflect that one allele is mutated in humans with TSC.

The Tsc2+/- mice have phenotypes that recapitulate ASD-like behavior in humans, such as impaired social interactions, learning, and synaptic plasticity. However, these mice (along with the Tsc1+/- mouse model) do not develop spontaneous benign tumor growth or seizures characteristic of the disorder TSC (von der Brelie, Waltereit et al. 2006,

Ehninger, Han et al. 2008, Auerbach, Osterweil et al. 2011, Potter, Basu et al. 2013,

Antoine, Langberg et al. 2019, Basu, Riordan et al. 2019). This highlights some of the challenges in using mice to model the highly variable disorder TSC and is suggestive of complex genetic and environmental interactions that factor into disease manifestation in humans. Alternative TSC mouse models include neuron-specific Tsc2 deletion and a hypomorph mouse where TSC2 levels are roughly 7% of wild-type mice (Yuan, Tsai et al. 2012). These mouse models have impaired social interaction and epilepsy characteristic of TSC, but it is unknown if they have an mGluR-LTD phenotype, which was a major focus of our work. Because work from patients and mouse models of TSC support a gene-dosage effect from loss of TSC1 and TSC2, such extreme mouse models may lead to conclusions not reflective of the underlying molecular dysfunction in TSC.

129 130

Another limitation is that TSC/mTOR signaling has cell-type specific effects that are missed by our whole tissue experimental approach. Methods to identify cell-type specific mRNA expression are RiboTag and Translating Ribosome Affinity Purification

(TRAP) (Sanz, Yang et al. 2009, Heiman, Kulicke et al. 2014). When combined with a cell-type specific Cre recombinase mouse, both methods allow for epitope tagging of a core ribosomal protein. mRNA-ribosome complexes from specific cells can be enriched by immunoprecipitation. RiboTag and TRAP could be used to identify changes in excitatory neurons from Tsc2+/- mice, but these methods lack the cell-type specific total mRNA comparison and thus, cannot distinguish if mRNA levels or translation drive differences in mRNA-ribosome association.

A goal of this study was to identify mRNAs that are dysregulated in Tsc2+/- mice following mGluR5 signaling, which was challenging given the mild changes and heterogeneity of the hippocampus. Because mGluR-LTD depends on protein synthesis in the neuropil (an area enriched for dendrites of CA1 pyramidal neurons) of the hippocampus, one approach is to micro-dissect the neuropil from the cell body of neurons. This approach was used by other laboratories to identify specific features in dendrites relative to the soma and has elucidated many unique aspects of dendritic neurobiology, such as enrichment of select mRNAs populations, alternative 3′UTR usage, and differential ribosome usage (Cajigas, Tushev et al. 2012, Tushev, Glock et al.

2018, Biever, Glock et al. 2020). Of note, micro-dissecting the neuropil from the soma requires a heroic effort to obtain sufficient material for RNA-seq and ribosome profiling.

130 131

As a perspective, our collective ribosome profiling and RNA-seq experiments utilized hippocampal slices with the cell body and dendrites from over 100 mice.

Concluding remarks

We initiated this study to identify the 1) mRNAs dysregulated in a mouse model of TSC autism and 2) responsive to induction of mGluR-LTD. Using RNA-seq and ribosome profiling, our work reveals that targets of FMRP are a common group of mRNAs that are elevated in TSC mice and rapidly increased following mGluR-LTD signaling. In addition, TSC mice have dysregulated eEF2K/eEF2 signaling to ribosome movement. Collectively, this work suggests a link between mGluR5-dependent eEF2K/eEF2 signaling, FMRP binding targets, and synaptic plasticity impairment in a mouse model of TSC autism. Future work is needed to determine if manipulating eEF2K/eEF2 signaling can rescue levels of FMRP binding targets and correct ASD-like behaviors in TSC mice.

Rapid protein synthesis is required for mGluR-LTD. We identified hundreds of mRNAs that rapidly increase and decrease in abundance following mGluR5 signaling.

Our work suggests that rapid remodeling of the synaptic proteome required for mGluR-

LTD likely includes mRNA turnover. Future work includes identifying possible regulators of mRNA degradation during mGluR-LTD, which may reveal novel mechanisms relevant to synaptic plasticity, higher cognitive function, and ASD-like behaviors.

131 132

REFERENCES

Acevedo, J. M., B. Hoermann, T. Schlimbach and A. A. Teleman (2018). "Changes in global translation elongation or initiation rates shape the proteome via the Kozak sequence." Sci Rep 8(1): 4018. Antion, M. D., L. Hou, H. Wong, C. A. Hoeffer and E. Klann (2008). "mGluR-dependent long-term depression is associated with increased phosphorylation of S6 and synthesis of elongation factor 1A but remains expressed in S6K-deficient mice." Mol Cell Biol 28(9): 2996-3007. Antoine, M. W., T. Langberg, P. Schnepel and D. E. Feldman (2019). "Increased Excitation-Inhibition Ratio Stabilizes Synapse and Circuit Excitability in Four Autism Mouse Models." Neuron 101(4): 648-661.e644. Ascano, M., N. Mukherjee, P. Bandaru, J. B. Miller, J. D. Nusbaum, D. L. Corcoran, C. Langlois, M. Munschauer, S. Dewell, M. Hafner, Z. Williams, U. Ohler and T. Tuschl (2012). "FMRP targets distinct mRNA sequence elements to regulate protein expression." Nature 492(7429): 382-386. Ashley, C. T., Jr., K. D. Wilkinson, D. Reines and S. T. Warren (1993). "FMR1 protein: conserved RNP family domains and selective RNA binding." Science 262(5133): 563-566. Astrinidis, A., T. P. Cash, D. S. Hunter, C. L. Walker, J. Chernoff and E. P. Henske (2002). "Tuberin, the tuberous sclerosis complex 2 tumor suppressor gene product, regulates Rho activation, cell adhesion and migration." Oncogene 21(55): 8470- 8476. Au, K. S., A. T. Williams, E. S. Roach, L. Batchelor, S. P. Sparagana, M. R. Delgado, J. W. Wheless, J. E. Baumgartner, B. B. Roa, C. M. Wilson, T. K. Smith-Knuppel, M. Y. Cheung, V. H. Whittemore, T. M. King and H. Northrup (2007). "Genotype/phenotype correlation in 325 individuals referred for a diagnosis of tuberous sclerosis complex in the United States." Genet Med 9(2): 88-100. Auerbach, B. D., E. K. Osterweil and M. F. Bear (2011). "Mutations causing syndromic autism define an axis of synaptic pathophysiology." Nature 480(7375): 63-68. Auerbach, B. D., E. K. Osterweil and M. F. Bear (2012). "Mutations causing syndromic autism define an axis of synaptic pathophysiology." Nature 480(7375): 63-68. Bagni, C., F. Tassone, G. Neri and R. Hagerman (2012). "Fragile X syndrome: causes, diagnosis, mechanisms, and therapeutics." J Clin Invest 122(12): 4314-4322. Bagni, C. and R. S. Zukin (2019). "A Synaptic Perspective of Fragile X Syndrome and Autism Spectrum Disorders." Neuron 101(6): 1070-1088. Banko, J. L., L. Hou, F. Poulin, N. Sonenberg and E. Klann (2006). "Regulation of eukaryotic initiation factor 4E by converging signaling pathways during metabotropic glutamate receptor-dependent long-term depression." J Neurosci 26(8): 2167-2173.

132 133

Banko, J. L., F. Poulin, L. Hou, C. T. DeMaria, N. Sonenberg and E. Klann (2005). "The translation repressor 4E-BP2 is critical for eIF4F complex formation, synaptic plasticity, and memory in the hippocampus." J Neurosci 25(42): 9581-9590. Basu, T., K. J. O. x. Riordan, B. A. Schoenike, N. N. Khan, E. P. Wallace, G. Rodriguez, R. K. Maganti and A. Roopra (2019). "Histone deacetylase inhibitors restore normal hippocampal synaptic plasticity and seizure threshold in a mouse model of Tuberous Sclerosis Complex." Scientific Reports: 1-11. Bateup, H. S., C. A. Johnson, C. L. Denefrio, J. L. Saulnier, K. Kornacker and B. L. Sabatini (2013). "Excitatory/Inhibitory Synaptic Imbalance Leads to Hippocampal Hyperexcitability in Mouse Models of Tuberous Sclerosis." Neuron 78(3): 510-522. Bateup, H. S., K. T. Takasaki, J. L. Saulnier, C. L. Denefrio and B. L. Sabatini (2011). "Loss of Tsc1 in vivo impairs hippocampal mGluR-LTD and increases excitatory synaptic function." J Neurosci 31(24): 8862-8869. Bear, M. F., K. M. Huber and S. T. Warren (2004). "The mGluR theory of fragile X mental retardation." Trends Neurosci 27(7): 370-377. Becker, A. H., E. Oh, J. S. Weissman, G. u. n. Kramer and B. Bukau (2013). "Selective ribosome profiling as a tool for studying the interaction of chaperones and targeting factors with nascent polypeptide chains and ribosomes." Nature protocols 8(11): 2212-2239. Berry-Kravis, E., V. Des Portes, R. Hagerman, S. Jacquemont, P. Charles, J. Visootsak, M. Brinkman, K. Rerat, B. Koumaras, L. Zhu, G. M. Barth, T. Jaecklin, G. Apostol and F. von Raison (2016). "Mavoglurant in fragile X syndrome: Results of two randomized, double-blind, placebo-controlled trials." Sci Transl Med 8(321): 321ra325. Bhattacharya, A., H. Kaphzan, A. C. Alvarez-Dieppa, J. P. Murphy, P. Pierre and E. Klann (2012). "Genetic Removal of p70 S6 Kinase 1 Corrects Molecular, Synaptic, and Behavioral Phenotypes in Fragile X Syndrome Mice." Neuron 76(2): 325-337. Biever, A., C. Glock, G. Tushev, E. Ciirdaeva, T. Dalmay, J. D. Langer and E. M. Schuman (2020). "Monosomes actively translate synaptic mRNAs in neuronal processes." Science 367(6477). Bilanges, B., R. Argonza-Barrett, M. Kolesnichenko, C. Skinner, M. Nair, M. Chen and D. Stokoe (2007). "Tuberous sclerosis complex proteins 1 and 2 control serum- dependent translation in a TOP-dependent and -independent manner." Mol Cell Biol 27(16): 5746-5764. Bontekoe, C. J., K. L. McIlwain, I. M. Nieuwenhuizen, L. A. Yuva-Paylor, A. Nellis, R. Willemsen, Z. Fang, L. Kirkpatrick, C. E. Bakker, R. McAninch, N. C. Cheng, M. Merriweather, A. T. Hoogeveen, D. Nelson, R. Paylor and B. A. Oostra (2002). "Knockout mouse model for Fxr2: a model for mental retardation." Hum Mol Genet 11(5): 487-498. Browne, G. J. and C. G. Proud (2004). "A novel mTOR-regulated phosphorylation site in elongation factor 2 kinase modulates the activity of the kinase and its binding to calmodulin." Mol Cell Biol 24(7): 2986-2997. Brunn, G. J., C. C. Hudson, A. Sekulic, J. M. Williams, H. Hosoi, P. J. Houghton, J. C. Lawrence, Jr. and R. T. Abraham (1997). "Phosphorylation of the translational

133 134

repressor PHAS-I by the mammalian target of rapamycin." Science 277(5322): 99- 101. Buxbaum and Singer (2014). "Single b-Actin mRNA Detection inNeurons Reveals a Mechanism forRegulating Its Translatability." Science 343(6169): 416-419. Cajigas, I. J., G. Tushev, T. J. Will, S. t. Dieck, N. Fuerst and E. M. Schuman (2012). "The Local Transcriptome in the Synaptic Neuropil Revealed by Deep Sequencing and High-Resolution Imaging." Neuron 74(3): 453-466. Carlberg, U., A. Nilsson and O. Nygard (1990). "Functional properties of phosphorylated elongation factor 2." Eur J Biochem 191(3): 639-645. Ceolin, L., N. Bouquier, J. Vitre-Boubaker, S. Rialle, D. Severac, E. Valjent, J. Perroy and E. Puighermanal (2017). "Cell Type-Specific mRNA Dysregulation in Hippocampal CA1 Pyramidal Neurons of the Fragile X Syndrome Mouse Model." Front Mol Neurosci 10: 340. Chan, L. Y., C. F. Mugler, S. Heinrich, P. Vallotton and K. Weis (2018). "Non-invasive measurement of mRNA decay reveals translation initiation as the major determinant of mRNA stability." Elife 7. Chang, H., J. Lim, M. Ha and V. N. Kim (2014). "TAIL-seq: genome-wide determination of poly(A) tail length and 3' end modifications." Mol Cell 53(6): 1044- 1052. Chang, H. H., P. H. Tsai, C. W. Liu, H. C. Ku, C. C. Kao and Y. H. Kao (2013). "Cycloheximide stimulates suppressor of cytokine signaling-3 gene expression in 3T3-L1 adipocytes via the extracellular signal-regulated kinase pathway." Toxicol Lett 217(1): 42-49. Chen, E., M. R. Sharma, X. Shi, R. K. Agrawal and S. Joseph (2014). "Fragile X Mental Retardation Protein Regulates Translation by Binding Directly to the Ribosome." Molecular Cell 54(3): 407-417. Chévere-Torres, I., H. Kaphzan, A. Bhattacharya, A. Kang, J. M. Maki, M. J. Gambello, J. L. Arbiser, E. Santini and E. Klann (2012). "Metabotropic glutamate receptor- dependent long-term depression is impaired due to elevated ERK signaling in the ΔRG mouse model of tuberous sclerosis complex." Neurobiol Dis 45(3): 1101-1110. Cho, J., N.-K. Yu, J.-H. Choi, S.-E. Sim, S. J. Kang, C. Kwak, S.-W. Lee, J.-i. Kim, D. I. Choi, V. N. Kim and B.-K. Kaang (2015). "Multiple repressive mechanisms in the hippocampus during memory formation." Science 350(6256): 82-87. Choi, Y. J., A. Di Nardo, I. Kramvis, L. Meikle, D. J. Kwiatkowski, M. Sahin and X. He (2008). "Tuberous sclerosis complex proteins control axon formation." Genes Dev 22(18): 2485-2495. Costa-Mattioli, M., D. Gobert, E. Stern, K. Gamache, R. Colina, C. Cuello, W. Sossin, R. Kaufman, J. Pelletier, K. Rosenblum, K. Krnjević, J. C. Lacaille, K. Nader and N. Sonenberg (2007). "eIF2alpha phosphorylation bidirectionally regulates the switch from short- to long-term synaptic plasticity and memory." Cell 129(1): 195-206. Darnell, J. C., S. J. Van Driesche, C. Zhang, K. Y. Hung, A. Mele, C. E. Fraser, E. F. Stone, C. Chen, J. J. Fak, S. W. Chi, D. D. Licatalosi, J. D. Richter and R. B. Darnell (2011).

134 135

"FMRP stalls ribosomal translocation on mRNAs linked to synaptic function and autism." Cell 146(2): 247-261. Darnell, J. C., S. J. Van Driesche, C. Zhang, K. Y. S. Hung, A. Mele, C. E. Fraser, E. F. Stone, C. Chen, J. J. Fak, S. W. Chi, D. D. Licatalosi, J. D. Richter and R. B. Darnell (2011). "FMRP Stalls Ribosomal Translocation on mRNAs Linked to Synaptic Function and Autism." Cell 146(2): 247-261. Darvishi, E. and G. M. Woldemichael (2016). "Cycloheximide Inhibits Actin Cytoskeletal Dynamics by Suppressing Signaling via RhoA." J Cell Biochem 117(12): 2886-2898. Dash, P. K., S. A. Orsi and A. N. Moore (2006). "Spatial memory formation and memory-enhancing effect of glucose involves activation of the tuberous sclerosis complex-Mammalian target of rapamycin pathway." J Neurosci 26(31): 8048-8056. Davidkova, G. and R. C. Carroll (2007). "Characterization of the role of microtubule- associated protein 1B in metabotropic glutamate receptor-mediated endocytosis of AMPA receptors in hippocampus." J Neurosci 27(48): 13273-13278. Di Nardo, A., M. H. Wertz, E. Kwiatkowski, P. T. Tsai, J. D. Leech, E. Greene-Colozzi, J. Goto, P. Dilsiz, D. M. Talos, C. B. Clish, D. J. Kwiatkowski and M. Sahin (2014). "Neuronal Tsc1/2 complex controls autophagy through AMPK-dependent regulation of ULK1." Hum Mol Genet 23(14): 3865-3874. Di Prisco, G. V., W. Huang, S. A. Buffington, C.-C. Hsu, P. E. Bonnen, A. N. Placzek, C. Sidrauski, K. Krnjević, R. J. Kaufman, P. Walter and M. Costa-Mattioli (2014). "Translational control of mGluR-dependent long-term depression and object-place learning by eIF2α." Nature Neuroscience 17(8): 1073-1082. Dölen, G., E. Osterweil, B. S. Rao, G. B. Smith, B. D. Auerbach, S. Chattarji and M. F. Bear (2007). "Correction of fragile X syndrome in mice." Neuron 56(6): 955-962. Dumont-Miscopein, A., J. P. Lavergne, D. Guillot, B. Sontag and J. P. Reboud (1994). "Interaction of phosphorylated elongation factor EF-2 with nucleotides and ribosomes." FEBS Lett 356(2-3): 283-286. Düvel, K., J. L. Yecies, S. Menon, P. Raman, A. I. Lipovsky, A. L. Souza, E. Triantafellow, Q. Ma, R. Gorski, S. Cleaver, M. G. Vander Heiden, J. P. MacKeigan, P. M. Finan, C. B. Clish, L. O. Murphy and B. D. Manning (2010). "Activation of a metabolic gene regulatory network downstream of mTOR complex 1." Mol Cell 39(2): 171-183. Ebrahimi-Fakhari, D., A. Saffari, L. Wahlster, A. Di Nardo, D. Turner, T. L. Lewis Jr., C. Conrad, J. M. Rothberg, J. O. Lipton, S. Kölker, G. F. Hoffmann, M.-J. Han, F. Polleux and M. Sahin (2016). "Impaired Mitochondrial Dynamics and Mitophagy in Neuronal Models of Tuberous Sclerosis Complex." Cell Reports 17(4): 1053-1070. Edbauer, D., J. R. Neilson, K. A. Foster, C. F. Wang, D. P. Seeburg, M. N. Batterton, T. Tada, B. M. Dolan, P. A. Sharp and M. Sheng (2010). "Regulation of synaptic structure and function by FMRP-associated microRNAs miR-125b and miR-132." Neuron 65(3): 373-384. Ehninger, D., S. Han, C. Shilyansky, Y. Zhou, W. Li, D. J. Kwiatkowski, V. Ramesh and A. J. Silva (2008). "Reversal of learning deficits in a Tsc2 +/− mouse model of tuberous sclerosis." Nature Medicine 14(8): 843-848.

135 136

Feng, Y., S. T. Warren and Warren (1997). "FMRP Associates with Polyribosomes as an mRNP, and the I304N Mutation of Severe Fragile X Syndrome Abolishes This Association." Molecular Cell: 1-10. Flexner, J. B., L. B. Flexner and E. Stellar (1963). "Memory in mice as affected by intracerebral puromycin." Science 141(3575): 57-59. Flexner, J. B., L. B. Flexner, E. Stellar, G. De La Haba and R. B. Roberts (1962). "Inhibition of protein synthesis in brain and learning and memory following puromycin." J Neurochem 9: 595-605. French, J. A., J. A. Lawson, Z. Yapici, H. Ikeda, T. Polster, R. Nabbout, P. Curatolo, P. J. de Vries, D. J. Dlugos, N. Berkowitz, M. Voi, S. Peyrard, D. Pelov and D. N. Franz (2016). "Adjunctive everolimus therapy for treatment-resistant focal-onset seizures associated with tuberous sclerosis (EXIST-3): a phase 3, randomised, double-blind, placebo-controlled study." Lancet 388(10056): 2153-2163. Fresno, M., A. Jiménez and D. Vázquez (1977). "Inhibition of translation in eukaryotic systems by harringtonine." Eur J Biochem 72(2): 323-330. Gandin, V., L. Masvidal, L. Hulea, S. P. Gravel, M. Cargnello, S. McLaughlan, Y. Cai, P. Balanathan, M. Morita, A. Rajakumar, L. Furic, M. Pollak, J. A. Porco, Jr., J. St-Pierre, J. Pelletier, O. Larsson and I. Topisirovic (2016). "nanoCAGE reveals 5' UTR features that define specific modes of translation of functionally related MTOR-sensitive mRNAs." Genome Res 26(5): 636-648. Garami, A., F. J. Zwartkruis, T. Nobukuni, M. Joaquin, M. Roccio, H. Stocker, S. C. Kozma, E. Hafen, J. L. Bos and G. Thomas (2003). "Insulin activation of Rheb, a mediator of mTOR/S6K/4E-BP signaling, is inhibited by TSC1 and 2." Mol Cell 11(6): 1457-1466. Garzia, A., S. M. Jafarnejad, C. Meyer, C. Chapat, T. Gogakos, P. Morozov, M. Amiri, M. Shapiro, H. Molina, T. Tuschl and N. Sonenberg (2017). "The E3 ubiquitin ligase and RNA-binding protein ZNF598 orchestrates ribosome quality control of premature polyadenylated mRNAs." Nat Commun 8: 16056. Gentilella, A., F. D. Moron-Duran, P. Fuentes, G. Zweig-Rocha, F. Riano-Canalias, J. Pelletier, M. Ruiz, G. Turon, J. Castano, A. Tauler, C. Bueno, P. Menendez, S. C. Kozma and G. Thomas (2017). "Autogenous Control of 5'TOP mRNA Stability by 40S Ribosomes." Mol Cell 67(1): 55-70.e54. Gildish, I., D. Manor, O. David, V. Sharma, D. Williams, U. Agarwala, X. Wang, J. W. Kenney, C. G. Proud and K. Rosenblum (2012). "Impaired associative taste learning and abnormal brain activation in kinase-defective eEF2K mice." Learn Mem 19(3): 116-125. Gingras, A. C., S. G. Kennedy, M. A. O'Leary, N. Sonenberg and N. Hay (1998). "4E- BP1, a repressor of mRNA translation, is phosphorylated and inactivated by the Akt(PKB) signaling pathway." Genes Dev 12(4): 502-513. Gismondi, A., S. Caldarola, G. Lisi, G. Juli, L. Chellini, V. Iadevaia, C. G. Proud and F. Loreni (2014). "Ribosomal stress activates eEF2K-eEF2 pathway causing translation elongation inhibition and recruitment of terminal oligopyrimidine (TOP) mRNAs on polysomes." Nucleic Acids Res 42(20): 12668-12680.

136 137

Graber, T. E., S. Hebert-Seropian, A. Khoutorsky, A. David, J. W. Yewdell, J. C. Lacaille and W. S. Sossin (2013). "Reactivation of stalled polyribosomes in synaptic plasticity." Proc Natl Acad Sci U S A 110(40): 16205-16210. Grabole, N., J. D. Zhang, S. Aigner, N. Ruderisch, V. Costa, F. C. Weber, M. Theron, N. Berntenis, O. Spleiss, M. Ebeling, G. W. Yeo, R. Jagasia and A. Kiialainen (2016). "Genomic analysis of the molecular neuropathology of tuberous sclerosis using a human stem cell model." Genome Medicine: 1-14. Gross, C., C. W. Chang, S. M. Kelly, A. Bhattacharya, S. M. McBride, S. W. Danielson, M. Q. Jiang, C. B. Chan, K. Ye, J. R. Gibson, E. Klann, T. A. Jongens, K. H. Moberg, K. M. Huber and G. J. Bassell (2015). "Increased expression of the PI3K enhancer PIKE mediates deficits in synaptic plasticity and behavior in fragile X syndrome." Cell Rep 11(5): 727-736. Guo, W., L. Zhang, D. M. Christopher, Z. Q. Teng, S. R. Fausett, C. Liu, O. L. George, J. Klingensmith, P. Jin and X. Zhao (2011). "RNA-binding protein FXR2 regulates adult hippocampal neurogenesis by reducing Noggin expression." Neuron 70(5): 924-938. Hafner, M., M. Landthaler, L. Burger, M. Khorshid, J. Hausser, P. Berninger, A. Rothballer, M. Ascano, Jr., A. C. Jungkamp, M. Munschauer, A. Ulrich, G. S. Wardle, S. Dewell, M. Zavolan and T. Tuschl (2010). "Transcriptome-wide identification of RNA-binding protein and microRNA target sites by PAR-CLIP." Cell 141(1): 129- 141. Hallett, L., T. Foster, Z. Liu, M. Blieden and J. Valentim (2011). "Burden of disease and unmet needs in tuberous sclerosis complex with neurological manifestations: systematic review." Curr Med Res Opin 27(8): 1571-1583. Heiman, M., R. Kulicke, R. J. Fenster, P. Greengard and N. Heintz (2014). "Cell type– specific mRNA purification by translating ribosome affinity purification (TRAP)." Nature protocols 9(6): 1282-1291. Heise, C., E. Taha, L. Murru, L. Ponzoni, A. Cattaneo, F. C. Guarnieri, C. Montani, A. Mossa, E. Vezzoli, G. Ippolito, J. Zapata, I. Barrera, A. G. Ryazanov, J. Cook, M. Poe, M. R. Stephen, M. Kopanitsa, R. Benfante, F. Rusconi, D. Braida, M. Francolini, C. G. Proud, F. Valtorta, M. Passafaro, M. Sala, A. Bachi, C. Verpelli, K. Rosenblum and C. Sala (2017). "eEF2K/eEF2 Pathway Controls the Excitation/Inhibition Balance and Susceptibility to Epileptic Seizures." Cereb Cortex 27(3): 2226-2248. Hernandez, O., S. Way, J. McKenna and M. J. Gambello (2007). "Generation of a conditional disruption of theTsc2 gene." genesis 45(2): 101-106. Hernandez, P. J. and T. Abel (2008). "The role of protein synthesis in memory consolidation: progress amid decades of debate." Neurobiol Learn Mem 89(3): 293- 311. Heyer, E. E., H. Ozadam, E. P. Ricci, C. Cenik and M. J. Moore (2015). "An optimized kit-free method for making strand-specific deep sequencing libraries from RNA fragments." Nucleic Acids Res 43(1): e2. Hinnebusch, A. G. (2005). "Translational regulation of GCN4 and the general amino acid control of yeast." Annu Rev Microbiol 59: 407-450.

137 138

Holt, C. E., K. C. Martin and E. M. Schuman (2019). "Local translation in neurons: visualization and function." Nat Struct Mol Biol 26(7): 557-566. Holz, M. K., B. A. Ballif, S. P. Gygi and J. Blenis (2005). "mTOR and S6K1 mediate assembly of the translation preinitiation complex through dynamic protein interchange and ordered phosphorylation events." Cell 123(4): 569-580. Hong, S., M. A. Freeberg, T. Han, A. Kamath, Y. Yao, T. Fukuda, T. Suzuki, J. K. Kim and K. Inoki (2017). "LARP1 functions as a molecular switch for mTORC1-mediated translation of an essential class of mRNAs." eLife 6: e25237-25224. Hou, L. and E. Klann (2004). "Activation of the phosphoinositide 3-kinase-Akt- mammalian target of rapamycin signaling pathway is required for metabotropic glutamate receptor-dependent long-term depression." J Neurosci 24(28): 6352- 6361. Huang, F., J. K. Chotiner and O. Steward (2005). "The mRNA for elongation factor 1alpha is localized in dendrites and translated in response to treatments that induce long-term depression." J Neurosci 25(31): 7199-7209. Huang, J., Y. Ikeuchi, M. Malumbres and A. Bonni (2015). "A Cdh1-APC/FMRP Ubiquitin Signaling Link Drives mGluR-Dependent Synaptic Plasticity in the Mammalian Brain." Neuron 86(3): 726-739. Huang, J. and B. D. Manning (2009). "A complex interplay between Akt, TSC2 and the two mTOR complexes." Biochem Soc Trans 37(Pt 1): 217-222. Huang, M. T. (1975). "Harringtonine, an inhibitor of initiation of protein biosynthesis." Mol Pharmacol 11(5): 511-519. Huber, K. M., S. M. Gallagher, S. T. Warren and M. F. Bear (2002). "Altered synaptic plasticity in a mouse model of fragile X mental retardation." Proc Natl Acad Sci U S A 99(11): 7746-7750. Huber, K. M., M. S. Kayser and M. F. Bear (2000). "Role for rapid dendritic protein synthesis in hippocampal mGluR-dependent long-term depression." Science 288(5469): 1254-1257. Hudson, C. C., M. Liu, G. G. Chiang, D. M. Otterness, D. C. Loomis, F. Kaper, A. J. Giaccia and R. T. Abraham (2002). "Regulation of hypoxia-inducible factor 1alpha expression and function by the mammalian target of rapamycin." Mol Cell Biol 22(20): 7004-7014. Ingolia, N. T., S. Ghaemmaghami, J. R. S. Newman and J. S. Weissman (2009). "Genome-wide analysis in vivo of translation with nucleotide resolution using ribosome profiling." Science 324(5924): 218-223. Inoki, K., Y. Li, T. Xu and K. L. Guan (2003). "Rheb GTPase is a direct target of TSC2 GAP activity and regulates mTOR signaling." Genes Dev 17(15): 1829-1834. Inoki, K., Y. Li, T. Zhu, J. Wu and K. L. Guan (2002). "TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling." Nat Cell Biol 4(9): 648-657. Inui, T., S. Kobayashi, Y. Ashikari, R. Sato, W. Endo, M. Uematsu, H. Oba, H. Saitsu, N. Matsumoto, S. Kure and K. Haginoya (2016). "Two cases of early-onset myoclonic seizures with continuous parietal delta activity caused by EEF1A2 mutations." Brain Dev 38(5): 520-524.

138 139

Iossifov, I., B. J. O'Roak, S. J. Sanders, M. Ronemus, N. Krumm, D. Levy, H. A. Stessman, K. T. Witherspoon, L. Vives, K. E. Patterson, J. D. Smith, B. Paeper, D. A. Nickerson, J. Dea, S. Dong, L. E. Gonzalez, J. D. Mandell, S. M. Mane, M. T. Murtha, C. A. Sullivan, M. F. Walker, Z. Waqar, L. Wei, A. J. Willsey, B. Yamrom, Y. H. Lee, E. Grabowska, E. Dalkic, Z. Wang, S. Marks, P. Andrews, A. Leotta, J. Kendall, I. Hakker, J. Rosenbaum, B. Ma, L. Rodgers, J. Troge, G. Narzisi, S. Yoon, M. C. Schatz, K. Ye, W. R. McCombie, J. Shendure, E. E. Eichler, M. W. State and M. Wigler (2014). "The contribution of de novo coding mutations to autism spectrum disorder." Nature 515(7526): 216-221. Ishizuka, A., M. C. Siomi and H. Siomi (2002). "A Drosophila fragile X protein interacts with components of RNAi and ribosomal proteins." Genes Dev 16(19): 2497-2508. Ishizuka, N., W. M. Cowan and D. G. Amaral (1995). "A quantitative analysis of the dendritic organization of pyramidal cells in the rat hippocampus." J Comp Neurol 362(1): 17-45. Jackson, R. J., C. U. Hellen and T. V. Pestova (2010). "The mechanism of eukaryotic translation initiation and principles of its regulation." Nat Rev Mol Cell Biol 11(2): 113-127. Jakkamsetti, V., N. P. Tsai, C. Gross, G. Molinaro, K. A. Collins, F. Nicoletti, K. H. Wang, P. Osten, G. J. Bassell, J. R. Gibson and K. M. Huber (2013). "Experience-induced Arc/Arg3.1 primes CA1 pyramidal neurons for metabotropic glutamate receptor- dependent long-term synaptic depression." Neuron 80(1): 72-79. Kang, H. and E. M. Schuman (1996). "A requirement for local protein synthesis in neurotrophin-induced hippocampal synaptic plasticity." Science 273(5280): 1402- 1406. Kelleher, R. J., 3rd and M. F. Bear (2008). "The autistic neuron: troubled translation?" Cell 135(3): 401-406. Kelly, E., S. M. Schaeffer, S. C. Dhamne, J. O. Lipton, L. Lindemann, M. Honer, G. Jaeschke, C. E. Super, S. H. Lammers, M. E. Modi, J. L. Silverman, J. R. Dreier, D. J. Kwiatkowski, A. Rotenberg and M. Sahin (2018). "mGluR5 Modulation of Behavioral and Epileptic Phenotypes in a Mouse Model of Tuberous Sclerosis Complex." Nature Publishing Group 43(6): 1457-1465. Khandjian, E. W., F. Corbin, S. Woerly and F. Rousseau (1996). "The fragile X mental retardation protein is associated with ribosomes." Nat Genet 12(1): 91-93. Kim, T. H., B. Tsang, R. M. Vernon, N. Sonenberg, L. E. Kay and J. D. Forman-Kay (2019). "Phospho-dependent phase separation of FMRP and CAPRIN1 recapitulates regulation of translation and deadenylation." Science 365(6455): 825-829. Korb, E., M. Herre, I. Zucker-Scharff, J. Gresack, C. D. Allis and R. B. Darnell (2017). "Excess Translation of Epigenetic Regulators Contributes to Fragile X Syndrome and Is Alleviated by Brd4 Inhibition." Cell 170(6): 1209-1223.e1220. Krol, J., V. Busskamp, I. Markiewicz, M. B. Stadler, S. Ribi, J. Richter, J. Duebel, S. Bicker, H. J. Fehling, D. Schübeler, T. G. Oertner, G. Schratt, M. Bibel, B. Roska and W. Filipowicz (2010). "Characterizing light-regulated retinal microRNAs reveals rapid turnover as a common property of neuronal microRNAs." Cell 141(4): 618-631.

139 140

Krueger, D. A., M. M. Care, K. Holland, K. Agricola, C. Tudor, P. Mangeshkar, K. A. Wilson, A. Byars, T. Sahmoud and D. N. Franz (2010). "Everolimus for subependymal giant-cell astrocytomas in tuberous sclerosis." N Engl J Med 363(19): 1801-1811. Krueger, D. A. and H. Northrup (2013). "Tuberous sclerosis complex surveillance and management: recommendations of the 2012 International Tuberous Sclerosis Complex Consensus Conference." Pediatr Neurol 49(4): 255-265. Krueger, D. A., A. Sadhwani, A. W. Byars, P. J. de Vries, D. N. Franz, V. H. Whittemore, R. Filip-Dhima, D. Murray, K. Kapur and M. Sahin (2017). "Everolimus for treatment of tuberous sclerosis complex-associated neuropsychiatric disorders." Ann Clin Transl Neurol 4(12): 877-887. Lamb, R. F., C. Roy, T. J. Diefenbach, H. V. Vinters, M. W. Johnson, D. G. Jay and A. Hall (2000). "The TSC1 tumour suppressor hamartin regulates cell adhesion through ERM proteins and the GTPase Rho." Nat Cell Biol 2(5): 281-287. Li, M., J. Shin, R. D. Risgaard, M. J. Parries, J. Wang, D. Chasman, S. Liu, S. Roy, A. Bhattacharyya and X. Zhao (2020). "Identification of FMR1-regulated molecular networks in human neurodevelopment." Genome Res 30(3): 361-374. Li, Y., M. E. Stockton, B. E. Eisinger, Y. Zhao, J. L. Miller, I. Bhuiyan, Y. Gao, Z. Wu, J. Peng and X. Zhao (2018). "Reducing histone acetylation rescues cognitive deficits in a mouse model of Fragile X syndrome." Nat Commun 9(1): 2494. Lipton, J. O. and M. Sahin (2014). "The Neurology of mTOR." Neuron 84(2): 275-291. Lüscher, C. and K. M. Huber (2010). "Group 1 mGluR-dependent synaptic long-term depression: mechanisms and implications for circuitry and disease." Neuron 65(4): 445-459. Martel, R. R., J. Klicius and S. Galet (1977). "Inhibition of the immune response by rapamycin, a new antifungal antibiotic." Can J Physiol Pharmacol 55(1): 48-51. Maurin, T., K. Lebrigand, S. Castagnola, A. Paquet, M. Jarjat, A. Popa, M. Grossi, F. Rage and B. Bardoni (2018). "HITS-CLIP in various brain areas reveals new targets and new modalities of RNA binding by fragile X mental retardation protein." Nucleic Acids Res 46(12): 6344-6355. McCamphill, P. K., C. A. Farah, M. N. Anadolu, S. Hoque and W. S. Sossin (2015). "Bidirectional Regulation of eEF2 Phosphorylation Controls Synaptic Plasticity by Decoding Neuronal Activity Patterns." Journal of Neuroscience 35(10): 4403-4417. McFleder, R. L., F. Mansur and J. D. Richter (2017). "Dynamic Control of Dendritic mRNA Expression by CNOT7 Regulates Synaptic Efficacy and Higher Cognitive Function." Cell Rep 20(3): 683-696. McManus, C. J., G. E. May, P. Spealman and A. Shteyman (2014). "Ribosome profiling reveals post-transcriptional buffering of divergent gene expression in yeast." Genome Res 24(3): 422-430. Miller, S., M. Yasuda, J. K. Coats, Y. Jones, M. E. Martone and M. Mayford (2002). "Disruption of dendritic translation of CaMKIIalpha impairs stabilization of synaptic plasticity and memory consolidation." Neuron 36(3): 507-519. Mills, E. W. and R. Green (2017). "Ribosomopathies: There’s strength in numbers." Science 358(6363): eaan2755-2710.

140 141

Morita, M., S.-P. Gravel, V. Chénard, K. Sikström, L. Zheng, T. Alain, V. Gandin, D. Avizonis, M. Arguello, C. Zakaria, S. McLaughlan, Y. Nouet, A. Pause, M. Pollak, E. Gottlieb, O. Larsson, J. St-Pierre, I. Topisirovic and N. Sonenberg (2013). "mTORC1 Controls Mitochondrial Activity and Biogenesis through 4E-BP-Dependent Translational Regulation." Cell Metabolism 18(5): 698-711. Muddashetty, R. S., V. C. Nalavadi, C. Gross, X. Yao, L. Xing, O. Laur, S. T. Warren and G. J. Bassell (2011). "Reversible inhibition of PSD-95 mRNA translation by miR-125a, FMRP phosphorylation, and mGluR signaling." Mol Cell 42(5): 673-688. Nakagawa, S. and M. E. Hauber (2011). "Great challenges with few subjects: Statistical strategies for neuroscientists." Neuroscience & Biobehavioral Reviews 35(3): 462-473. Nakajima, J., N. Okamoto, J. Tohyama, M. Kato, H. Arai, O. Funahashi, Y. Tsurusaki, M. Nakashima, H. Kawashima, H. Saitsu, N. Matsumoto and N. Miyake (2015). "De novo EEF1A2 mutations in patients with characteristic facial features, intellectual disability, autistic behaviors and epilepsy." Clin Genet 87(4): 356-361. Nalavadi, V. C., R. S. Muddashetty, C. Gross and G. J. Bassell (2012). "Dephosphorylation-induced ubiquitination and degradation of FMRP in dendrites: a role in immediate early mGluR-stimulated translation." J Neurosci 32(8): 2582- 2587. Nalavadi, V. C., R. S. Muddashetty, C. Gross and G. J. Bassell (2012). "Dephosphorylation-induced ubiquitination and degradation of FMRP in dendrites: a role in immediate early mGluR-stimulated translation." The Journal of neuroscience : the official journal of the Society for Neuroscience 32(8): 2582-2587. Nelson, S. B. and V. Valakh (2015). "Excitatory/Inhibitory Balance and Circuit Homeostasis in Autism Spectrum Disorders." Neuron 87(4): 684-698. Neuman, N. A. and E. P. Henske (2011). "Non-canonical functions of the tuberous sclerosis complex-Rheb signalling axis." EMBO Mol Med 3(4): 189-200. Nie, D., A. Di Nardo, J. M. Han, H. Baharanyi, I. Kramvis, T. Huynh, S. Dabora, S. Codeluppi, P. P. Pandolfi, E. B. Pasquale and M. Sahin (2010). "Tsc2-Rheb signaling regulates EphA-mediated axon guidance." Nat Neurosci 13(2): 163-172. Niere, F., J. R. Wilkerson and K. M. Huber (2012). "Evidence for a fragile X mental retardation protein-mediated translational switch in metabotropic glutamate receptor-triggered Arc translation and long-term depression." J Neurosci 32(17): 5924-5936. Nilsson, L. and O. Nygård (1991). "Altered sensitivity of eukaryotic elongation factor 2 for trypsin after phosphorylation and ribosomal binding." J Biol Chem 266(16): 10578-10582. Northrup, H., D. J. Kwiatkowski, E. S. Roach, W. B. Dobyns, R. A. Lewis, G. E. Herman, E. Rodriguez, Jr., S. P. Daiger and S. H. Blanton (1992). "Evidence for genetic heterogeneity in tuberous sclerosis: one locus on chromosome 9 and at least one locus elsewhere." Am J Hum Genet 51(4): 709-720. O'Roak, B. J., L. Vives, S. Girirajan, E. Karakoc, N. Krumm, B. P. Coe, R. Levy, A. Ko, C. Lee, J. D. Smith, E. H. Turner, I. B. Stanaway, B. Vernot, M. Malig, C. Baker, B. Reilly, J.

141 142

M. Akey, E. Borenstein, M. J. Rieder, D. A. Nickerson, R. Bernier, J. Shendure and E. E. Eichler (2012). "Sporadic autism exomes reveal a highly interconnected protein network of de novo mutations." Nature 485(7397): 246-250. Oberlé, I., F. Rousseau, D. Heitz, C. Kretz, D. Devys, A. Hanauer, J. Boué, M. F. Bertheas and J. L. Mandel (1991). "Instability of a 550-base pair DNA segment and abnormal methylation in fragile X syndrome." Science 252(5009): 1097-1102. Oh, E., A. H. Becker, A. Sandikci, D. Huber, R. Chaba, F. Gloge, R. J. Nichols, A. Typas, C. A. Gross, G. Kramer, J. S. Weissman and B. Bukau (2011). "Selective Ribosome Profiling Reveals the Cotranslational Chaperone Action of Trigger Factor In Vivo." Cell 147(6): 1295-1308. Oksvold, M. P., N. M. Pedersen, L. Forfang and E. B. Smeland (2012). "Effect of cycloheximide on epidermal growth factor receptor trafficking and signaling." FEBS Lett 586(20): 3575-3581. Olde Loohuis, N. F., W. Ba, P. H. Stoerchel, A. Kos, A. Jager, G. Schratt, G. J. Martens, H. van Bokhoven, N. Nadif Kasri and A. Aschrafi (2015). "MicroRNA-137 Controls AMPA-Receptor-Mediated Transmission and mGluR-Dependent LTD." Cell Rep 11(12): 1876-1884. Ouwenga, R. L. (2015). "Fmrp targets or not: long, highly brain-expressed genes tend to be implicated in autism and brain disorders." 1-7. Overwater, I. E., A. B. Rietman, S. E. Mous, K. Bindels-de Heus, D. Rizopoulos, L. W. Ten Hoopen, T. van der Vaart, F. E. Jansen, Y. Elgersma, H. A. Moll and M. Y. de Wit (2019). "A randomized controlled trial with everolimus for IQ and autism in tuberous sclerosis complex." Neurology 93(2): e200-e209. Park, S., J. M. Park, S. Kim, J. A. Kim, J. D. Shepherd, C. L. Smith-Hicks, S. Chowdhury, W. Kaufmann, D. Kuhl, A. G. Ryazanov, R. L. Huganir, D. J. Linden and P. F. Worley (2008). "Elongation factor 2 and fragile X mental retardation protein control the dynamic translation of Arc/Arg3.1 essential for mGluR-LTD." Neuron 59(1): 70-83. Peña-Llopis, S., S. Vega-Rubin-de-Celis, J. C. Schwartz, N. C. Wolff, T. A. Tran, L. Zou, X. J. Xie, D. R. Corey and J. Brugarolas (2011). "Regulation of TFEB and V-ATPases by mTORC1." Embo j 30(16): 3242-3258. Peters, J. M., R. R. Struyven, A. K. Prohl, L. Vasung, A. Stajduhar, M. Taquet, J. J. Bushman, H. Lidov, J. M. Singh, B. Scherrer, J. R. Madsen, S. P. Prabhu, M. Sahin, O. Afacan and S. K. Warfield (2019). "White matter mean diffusivity correlates with myelination in tuberous sclerosis complex." Ann Clin Transl Neurol 6(7): 1178- 1190. Philippe, L., J.-J. Vasseur, F. Debart and C. C. Thoreen (2017). "La-related protein 1 (LARP1) repression of TOP mRNA translation is mediated through its cap-binding domain and controlled by an adjacent regulatory region." Nucleic Acids Research. Pieretti, M., F. P. Zhang, Y. H. Fu, S. T. Warren, B. A. Oostra, C. T. Caskey and D. L. Nelson (1991). "Absence of expression of the FMR-1 gene in fragile X syndrome." Cell 66(4): 817-822. Potter, W. B., T. Basu, K. J. O'Riordan, A. Kirchner, P. Rutecki, C. Burger and A. Roopra (2013). "Reduced juvenile long-term depression in tuberous sclerosis complex is

142 143

mitigated in adults by compensatory recruitment of mGluR5 and Erk signaling." PLoS Biol 11(8): e1001627. Price, N. T., N. T. Redpath, K. V. Severinov, D. G. Campbell, J. M. Russell and C. G. Proud (1991). "Identification of the phosphorylation sites in elongation factor-2 from rabbit reticulocytes." FEBS Lett 282(2): 253-258. Rachid El Fatimy, L. D. S. T. X. J. A. D. C. R. P. D. K. E. W. K. (2016). "Tracking the Fragile X Mental Retardation Protein in a Highly Ordered Neuronal RiboNucleoParticles Population: A Link between Stalled Polyribosomes and RNA Granules." 1-31. Raught, B., F. Peiretti, A. C. Gingras, M. Livingstone, D. Shahbazian, G. L. Mayeur, R. D. Polakiewicz, N. Sonenberg and J. W. Hershey (2004). "Phosphorylation of eucaryotic translation initiation factor 4B Ser422 is modulated by S6 kinases." Embo j 23(8): 1761-1769. Redpath, N. T., E. J. Foulstone and C. G. Proud (1996). "Regulation of translation elongation factor-2 by insulin via a rapamycin-sensitive signalling pathway." Embo j 15(9): 2291-2297. Richards, C., C. Jones, L. Groves, J. Moss and C. Oliver (2015). "Prevalence of autism spectrum disorder phenomenology in genetic disorders: a systematic review and meta-analysis." Lancet Psychiatry 2(10): 909-916. Richter, J. D. and J. Coller (2015). "Pausing on Polyribosomes: Make Way for Elongation in Translational Control." Cell 163(2): 292-300. Ridler, K., J. Suckling, N. Higgins, P. Bolton and E. Bullmore (2004). "Standardized whole brain mapping of tubers and subependymal nodules in tuberous sclerosis complex." J Child Neurol 19(9): 658-665. Ronesi, J. A., K. A. Collins, S. A. Hays, N.-P. Tsai, W. Guo, S. G. Birnbaum, J.-H. Hu, P. F. Worley, J. R. Gibson and K. M. Huber (2012). "Disrupted Homer scaffolds mediate abnormal mGluR5 function in a mouse model of fragile X syndrome." Nature neuroscience 15(3): 431-440- S431. Rosset, C., C. B. O. Netto and P. Ashton-Prolla (2017). "TSC1 and TSC2 gene mutations and their implications for treatment in Tuberous Sclerosis Complex: a review." Genet Mol Biol 40(1): 69-79. Roux, P. P. and I. Topisirovic (2018). "Signaling Pathways Involved in the Regulation of mRNA Translation." Mol Cell Biol 38(12). Ryazanov, A. G., E. A. Shestakova and P. G. Natapov (1988). "Phosphorylation of elongation factor 2 by EF-2 kinase affects rate of translation." Nature 334(6178): 170-173. Sadowski, K., K. Kotulska, R. A. Schwartz and S. Jozwiak (2016). "Systemic effects of treatment with mTOR inhibitors in tuberous sclerosis complex: a comprehensive review." J Eur Acad Dermatol Venereol 30(4): 586-594. Sampson, J. R. and P. C. Harris (1994). "The molecular genetics of tuberous sclerosis." Hum Mol Genet 3 Spec No: 1477-1480.

143 144

Sanz, E., L. Yang, T. Su, D. R. Morris, G. S. McKnight and P. S. Amieux (2009). "Cell- type-specific isolation of ribosome-associated mRNA from complex tissues." Proc Natl Acad Sci U S A 106(33): 13939-13944. Sarbassov, D. D., S. M. Ali, S. Sengupta, J. H. Sheen, P. P. Hsu, A. F. Bagley, A. L. Markhard and D. M. Sabatini (2006). "Prolonged rapamycin treatment inhibits mTORC2 assembly and Akt/PKB." Mol Cell 22(2): 159-168. Saré, R. M., C. Figueroa, A. Lemons, I. Loutaev and C. Beebe Smith (2019). "Comparative Behavioral Phenotypes of Fmr1 KO, Fxr2 Het, and Fmr1 KO/Fxr2 Het Mice." Brain Sci 9(1). Sato, A., S. Kasai, T. Kobayashi, Y. Takamatsu, O. Hino, K. Ikeda and M. Mizuguchi (2019). "Rapamycin reverses impaired social interaction in mouse models of tuberous sclerosis complex." 1-9. Sawicka, K., C. R. Hale, C. Y. Park, J. J. Fak, J. E. Gresack, S. J. Van Driesche, J. J. Kang, J. C. Darnell and R. B. Darnell (2019). "FMRP has a cell-type-specific role in CA1 pyramidal neurons to regulate autism-related transcripts and circadian memory." Elife 8. Saxton, R. A. and D. M. Sabatini (2017). "mTOR Signaling in Growth, Metabolism, and Disease." Cell 168(6): 960-976. Schaeffer, C., B. Bardoni, J. L. Mandel, B. Ehresmann, C. Ehresmann and H. Moine (2001). "The fragile X mental retardation protein binds specifically to its mRNA via a purine quartet motif." Embo j 20(17): 4803-4813. Scheetz, A. J., A. C. Nairn and M. Constantine-Paton (2000). "NMDA receptor- mediated control of protein synthesis at developing synapses." Nat Neurosci 3(3): 211-216. Schneider-Poetsch, T., J. Ju, D. E. Eyler, Y. Dang, S. Bhat, W. C. Merrick, R. Green, B. Shen and J. O. Liu (2010). "Inhibition of eukaryotic translation elongation by cycloheximide and lactimidomycin." Nat Chem Biol 6(3): 209-217. Settembre, C., R. Zoncu, D. L. Medina, F. Vetrini, S. Erdin, S. Erdin, T. Huynh, M. Ferron, G. Karsenty, M. C. Vellard, V. Facchinetti, D. M. Sabatini and A. Ballabio (2012). "A lysosome-to-nucleus signalling mechanism senses and regulates the lysosome via mTOR and TFEB." Embo j 31(5): 1095-1108. Shah, S., G. Molinaro, B. Liu, R. Wang, K. M. Huber and J. D. Richter (2020). "FMRP Control of Ribosome Translocation Promotes Chromatin Modifications and Alternative Splicing of Neuronal Genes Linked to Autism." Cell Reports (in press). Sharma, A., C. A. Hoeffer, Y. Takayasu, T. Miyawaki, S. M. McBride, E. Klann and R. S. Zukin (2010). "Dysregulation of mTOR signaling in fragile X syndrome." J Neurosci 30(2): 694-702. Shaw, R. J. and L. C. Cantley (2006). "Ras, PI(3)K and mTOR signalling controls tumour cell growth." Nature 441(7092): 424-430. Shu, H., E. Donnard, B. Liu and J. D. Richter (2019). "Genetic Rescue of Fragile X Syndrome Links FMRP Deficiency to Codon Optimality-Dependent RNA Destabilization." bioRxiv.

144 145

Shu, H., E. Donnard, B. Liu, R. Wang and J. D. Richter (2020). "FMRP Links Optimal Codons to mRNA stability in Neurons." bioRxiv: 801449. Singh, K., S. Sun and C. Vézina (1979). "Rapamycin (AY-22,989), a new antifungal antibiotic. IV. Mechanism of action." J Antibiot (Tokyo) 32(6): 630-645. Siomi, M. C., Y. Zhang, H. Siomi and G. Dreyfuss (1996). "Specific sequences in the fragile X syndrome protein FMR1 and the FXR proteins mediate their binding to 60S ribosomal subunits and the interactions among them." Mol Cell Biol 16(7): 3825- 3832. Smalley, S. L., F. Burger and M. Smith (1994). "Phenotypic variation of tuberous sclerosis in a single extended kindred." J Med Genet 31(10): 761-765. Sossin, W. S. and M. Costa-Mattioli (2019). "Translational Control in the Brain in Health and Disease." Cold Spring Harb Perspect Biol 11(8). Spencer, C. M., E. Serysheva, L. A. Yuva-Paylor, B. A. Oostra, D. L. Nelson and R. Paylor (2006). "Exaggerated behavioral phenotypes in Fmr1/Fxr2 double knockout mice reveal a functional genetic interaction between Fragile X-related proteins." Hum Mol Genet 15(12): 1984-1994. Stefani, G. (2004). "Fragile X Mental Retardation Protein Is Associated with Translating Polyribosomes in Neuronal Cells." Journal of Neuroscience 24(33): 7272-7276. Stefani, G., C. E. Fraser, J. C. Darnell and R. B. Darnell (2004). "Fragile X mental retardation protein is associated with translating polyribosomes in neuronal cells." J Neurosci 24(33): 7272-7276. Sutton, M. A., A. M. Taylor, H. T. Ito, A. Pham and E. M. Schuman (2007). "Postsynaptic decoding of neural activity: eEF2 as a biochemical sensor coupling miniature synaptic transmission to local protein synthesis." Neuron 55(4): 648-661. Svitkin, Y. V., A. Pause, A. Haghighat, S. Pyronnet, G. Witherell, G. J. Belsham and N. Sonenberg (2001). "The requirement for eukaryotic initiation factor 4A (elF4A) in translation is in direct proportion to the degree of mRNA 5' secondary structure." Rna 7(3): 382-394. Tabet, R., E. Moutin, J. A. Becker, D. Heintz, L. Fouillen, E. Flatter, W. Krężel, V. Alunni, P. Koebel, D. Dembélé, F. Tassone, B. Bardoni, J. L. Mandel, N. Vitale, D. Muller, J. Le Merrer and H. Moine (2016). "Fragile X Mental Retardation Protein (FMRP) controls diacylglycerol kinase activity in neurons." Proc Natl Acad Sci U S A 113(26): E3619- 3628. Tang, S. J., G. Reis, H. Kang, A. C. Gingras, N. Sonenberg and E. M. Schuman (2002). "A rapamycin-sensitive signaling pathway contributes to long-term synaptic plasticity in the hippocampus." Proc Natl Acad Sci U S A 99(1): 467-472. Tavazoie, S. F., V. A. Alvarez, D. A. Ridenour, D. J. Kwiatkowski and B. L. Sabatini (2005). "Regulation of neuronal morphology and function by the tumor suppressors Tsc1 and Tsc2." Nat Neurosci 8(12): 1727-1734. Thoreen, C. C., L. Chantranupong, H. R. Keys, T. Wang, N. S. Gray and D. M. Sabatini (2012). "A unifying model for mTORC1-mediated regulation of mRNA translation." Nature 486(7396): 109-113.

145 146

Tsai, P. and M. Sahin (2011). "Mechanisms of neurocognitive dysfunction and therapeutic considerations in tuberous sclerosis complex." Curr Opin Neurol 24(2): 106-113. Tsang, B., J. Arsenault, R. M. Vernon, H. Lin, N. Sonenberg, L. Y. Wang, A. Bah and J. D. Forman-Kay (2019). "Phosphoregulated FMRP phase separation models activity- dependent translation through bidirectional control of mRNA granule formation." Proc Natl Acad Sci U S A 116(10): 4218-4227. Tushev, G., C. Glock, M. Heumüller, A. Biever, M. Jovanovic and E. M. Schuman (2018). "Alternative 3' UTRs Modify the Localization, Regulatory Potential, Stability, and Plasticity of mRNAs in Neuronal Compartments." Neuron 98(3): 495-511.e496. Udagawa, T., N. G. Farny, M. Jakovcevski, H. Kaphzan, J. M. Alarcon, S. Anilkumar, M. Ivshina, J. A. Hurt, K. Nagaoka, V. C. Nalavadi, L. J. Lorenz, G. J. Bassell, S. Akbarian, S. Chattarji, E. Klann and J. D. Richter (2013). "Genetic and acute CPEB1 depletion ameliorate fragile X pathophysiology." Nature Medicine 19(11): 1473-1477. van Slegtenhorst, M., R. de Hoogt, C. Hermans, M. Nellist, B. Janssen, S. Verhoef, D. Lindhout, A. van den Ouweland, D. Halley, J. Young, M. Burley, S. Jeremiah, K. Woodward, J. Nahmias, M. Fox, R. Ekong, J. Osborne, J. Wolfe, S. Povey, R. G. Snell, J. P. Cheadle, A. C. Jones, M. Tachataki, D. Ravine, J. R. Sampson, M. P. Reeve, P. Richardson, F. Wilmer, C. Munro, T. L. Hawkins, T. Sepp, J. B. Ali, S. Ward, A. J. Green, J. R. Yates, J. Kwiatkowska, E. P. Henske, M. P. Short, J. H. Haines, S. Jozwiak and D. J. Kwiatkowski (1997). "Identification of the tuberous sclerosis gene TSC1 on chromosome 9q34." Science 277(5327): 805-808. Verkerk, A. J., M. Pieretti, J. S. Sutcliffe, Y. H. Fu, D. P. Kuhl, A. Pizzuti, O. Reiner, S. Richards, M. F. Victoria, F. P. Zhang and et al. (1991). "Identification of a gene (FMR- 1) containing a CGG repeat coincident with a breakpoint cluster region exhibiting length variation in fragile X syndrome." Cell 65(5): 905-914. Verpelli, C., G. Piccoli, C. Zibetti, A. Zanchi, F. Gardoni, K. Huang, D. Brambilla, M. Di Luca, E. Battaglioli and C. Sala (2010). "Synaptic activity controls dendritic spine morphology by modulating eEF2-dependent BDNF synthesis." J Neurosci 30(17): 5830-5842. Vézina, C., A. Kudelski and S. N. Sehgal (1975). "Rapamycin (AY-22,989), a new antifungal antibiotic. I. Taxonomy of the producing streptomycete and isolation of the active principle." J Antibiot (Tokyo) 28(10): 721-726. von der Brelie, C., R. Waltereit, L. Zhang, H. Beck and T. Kirschstein (2006). "Impaired synaptic plasticity in a rat model of tuberous sclerosis." Eur J Neurosci 23(3): 686-692. Walden, W. E. and R. E. Thach (1986). "Translational control of gene expression in a normal fibroblast. Characterization of a subclass of mRNAs with unusual kinetic properties." Biochemistry 25(8): 2033-2041. Wang, G. X., S. J. Smith and P. Mourrain (2016). "Sub-synaptic, multiplexed analysis of proteins reveals Fragile X related protein 2 is mislocalized in Fmr1 KO synapses." Elife 5.

146 147

Wang, X., W. Li, M. Williams, N. Terada, D. R. Alessi and C. G. Proud (2001). "Regulation of elongation factor 2 kinase by p90(RSK1) and p70 S6 kinase." Embo j 20(16): 4370-4379. Waung, M. W. and K. M. Huber (2009). "Protein translation in synaptic plasticity: mGluR-LTD, Fragile X." Curr Opin Neurobiol 19(3): 319-326. Waung, M. W., B. E. Pfeiffer, E. D. Nosyreva, J. A. Ronesi and K. M. Huber (2008). "Rapid translation of Arc/Arg3.1 selectively mediates mGluR-dependent LTD through persistent increases in AMPAR endocytosis rate." Neuron 59(1): 84-97. Weng, W., Y. Chen, M. Wang, Y. Zhuang and T. Behnisch (2016). "Potentiation of Schaffer-Collateral CA1 Synaptic Transmission by eEF2K and p38 MAPK Mediated Mechanisms." Front Cell Neurosci 10: 247. Wong, V. and P. L. Khong (2006). "Tuberous sclerosis complex: correlation of magnetic resonance imaging (MRI) findings with comorbidities." J Child Neurol 21(2): 99-105. Yang, Q., K. Inoki, E. Kim and K. L. Guan (2006). "TSC1/TSC2 and Rheb have different effects on TORC1 and TORC2 activity." Proc Natl Acad Sci U S A 103(18): 6811-6816. Yu, G., L. G. Wang, Y. Han and Q. Y. He (2012). "clusterProfiler: an R package for comparing biological themes among gene clusters." Omics 16(5): 284-287. Yuan, E., P. T. Tsai, E. Greene-Colozzi, M. Sahin, D. J. Kwiatkowski and I. A. Malinowska (2012). "Graded loss of tuberin in an allelic series of brain models of TSC correlates with survival, and biochemical, histological and behavioral features." Hum Mol Genet 21(19): 4286-4300. Zalfa, F., B. Eleuteri, K. S. Dickson, V. Mercaldo, S. De Rubeis, A. di Penta, E. Tabolacci, P. Chiurazzi, G. Neri, S. G. Grant and C. Bagni (2007). "A new function for the fragile X mental retardation protein in regulation of PSD-95 mRNA stability." Nat Neurosci 10(5): 578-587. Zhang, F., Y. Kang, M. Wang, Y. Li, T. Xu, W. Yang, H. Song, H. Wu, Q. Shu and P. Jin (2018). "Fragile X mental retardation protein modulates the stability of its m6A- marked messenger RNA targets." Hum Mol Genet 27(22): 3936-3950. Zhang, Y., X. Gao, L. J. Saucedo, B. Ru, B. A. Edgar and D. Pan (2003). "Rheb is a direct target of the tuberous sclerosis tumour suppressor proteins." Nat Cell Biol 5(6): 578-581. Zhu, P. J., C. J. Chen, J. Mays, L. Stoica and M. Costa-Mattioli (2018). "mTORC2, but not mTORC1, is required for hippocampal mGluR-LTD and associated behaviors." Nat Neurosci 21(6): 799-802.

147