Gene Therapy for Facioscapulohumeral Muscular Dystrophy

Dissertation

Presented in Partial Fulfillment of the Requirements for

the Degree Doctor of Philosophy in the Graduate

School of The Ohio State University

By

Lindsay M. Wallace, B.A.

Molecular, Cellular, and Developmental Biology Graduate Program

The Ohio State University

2012

Dissertation Committee:

Dr. Scott Q. Harper, Advisor

Dr. K. Reed Clark

Dr. Denis C. Guttridge

Dr. Jill Rafael-Fortney

1

Copyright by

Lindsay M. Wallace

2012

ABSTRACT

Facioscapulohumeral muscular dystrophy is the most common inherited muscular

dystrophy though no treatment exists. The lack of therapeutic development for FSHD is

directly linked to insufficient understanding of how the disease is caused. The goals of

the studies presented here were to gain a better understanding of the pathogenic

mechanisms of FSHD and to develop targeted translational therapies to treat the disease.

FSHD is associated with D4Z4 repeat contraction on human chromosome 4q35,

which does not result in complete loss or mutation of any gene. Consequently, the major

obstacle to discerning the underlying pathogenic mechanism is to identify the cause.

Although no gene was conclusively linked to FSHD development, evidence supported a

role for the D4Z4-encoded DUX4 gene. In Chapter 3, our objective was to test the in

vivo myopathic potential of DUX4. We delivered DUX4 to zebrafish and mouse muscle

by transposon-mediated transgenesis and adeno-associated viral vectors, respectively. We

found over-expression of DUX4 caused abnormalities associated with muscular

dystrophy in both animal models. This toxicity required DNA binding, since a DUX4

DNA binding domain mutant produced no abnormalities. We also found the toxic effects of DUX4 were p53-dependent. This study demonstrated the myopathic potential of

DUX4 in animal muscle and provided a p53-dependent mechanism for DUX4-induced

ii toxicity. Considering previous studies showed DUX4 was elevated in FSHD patient

muscles, our data support the hypothesis that DUX4 over-expression contributes to

FSHD development.

With DUX4 as a potential target, gene silencing approaches could provide treatment for FSHD. With as many as 29 different gene mutations responsible for other

dominant myopathies, gene silencing approaches could have a broad impact. Feasible

mechanisms to silence dominant disease genes have lagged behind gene replacement

strategies, but with the discovery of RNA interference (RNAi) and its subsequent

development into a promising new gene silencing tool, the landscape has changed. In

Chapter 4, our objective was to demonstrate proof-of-principle for RNAi therapy of a

dominant myopathy in vivo. We tested the potential of AAV-delivered therapeutic microRNAs, targeting the human Facioscapulohumeral muscular dystrophy (FSHD)

Region Gene 1 (FRG1), to correct myopathic features in mice expressing toxic levels of human FRG1 (FRG1-high mice). We found that FRG1 gene silencing improved muscle

mass, strength, and histopathological abnormalities associated with muscular dystrophy

in FRG1-high mice, thereby demonstrating therapeutic promise for treatment of FSHD and

other dominantly inherited myopathies using RNAi.

Next we applied this therapeutic strategy to FSHD by targeting DUX4. Several

recent studies support an FSHD pathogenesis model involving over-expression of the myopathic DUX4 gene making it the most promising therapeutic target. In Chapter 5, we tested a pre-clinical RNAi-based DUX4 gene silencing approach as a prospective treatment for FSHD. We found that AAV vector-delivered therapeutic microRNAs corrected DUX4-associated myopathy in mouse muscle. These results provide proof-of-

iii principle for RNAi therapy of FSHD through DUX4 inhibition. Together these studies have helped define the main pathogenic insult in FSHD and laid out a plausible, targeted therapy to treat the disease.

iv

ACKNOWLEDGMENTS

First and foremost I would like to thank my mentor and advisor, Dr. Scott Harper.

I do not have enough words to express all he has done for me over the years. He truly embodies both of the titles Mentor and Advisor, and I would not be where I am today if not for him. He has provided helpful advice and immense opportunities as well as unwavering confidence and support during my time with him. He encouraged me to think big and gave me all the tools to achieve my goals. He acknowledges my strengths and helps me work through my weaknesses. My professional success as a scientist is a direct reflection of his encouragement and support and for that I am truly thankful.

I thank my committee members Dr. Denis Guttridge, Dr. Reed Clark, and Dr. Jill

Rafael-Fortney for all of their helpful discussions in committee meetings, muscle group meetings and journal clubs. They have each provided a different perspective on my projects, the field of science in general, and how to be successful in research. Because of them I received a diverse and comprehensive education. I value what all members have taught me and thank them for pushing me to my full potential.

I would also like to acknowledge those from the beginning of my scientific career.

I thank Dr. Mike Robinson for taking a chance on me when I was an inexperienced undergraduate student. He introduced me to the world of research, and got me excited about science. Also thanks to Dr. Heithem El-Hodiri, who convinced me to apply to

v graduate school. Without his support and encouragement I would not be in this field. I would also like to thank him for his continued support and advice during my first few years of graduate school; he has been a trusted confidant and career advisor.

I would like to thank all members of the Harper Lab for helping me reach this goal. They have served as a support system both scientifically and personally, especially

Sara Garwick-Coppens who has been there with me from the very beginning. I would also like to thank friends from my graduate program: Matt Murtha, Tom Bebee, Dan

Deathridge, Kevin Bosse and Brian Hutzen and well as my roommates and IBGP girl friends: Amanda (Haidet) Phillips, Julie (Johnson) Moravy, Sasha Beyer, Stephanie

Cush, and Amanda (Beltramini) Healan for helping me through this process and for being there during all the tough times of graduate school.

Finally, I want to thank my family; (parents) Phil and Jeri Wallace, (sister) Laura

(Wallace) Harrington, and (grandparents) Jim and Lois Chamberlain and Dick and

Sharon Wallace for always believing in me, encouraging education and giving me everything I have needed in life to be successful. I also want to thank my fiancé, Eric

Winner, for his never-ending support both personally and technically. He has literally stood by my side through late night experiments, weekend vivarium runs, last minute poster printings and every Harper Lab computer problem that has ever existed. On top of that, he has truly encouraged me to be the best person I can be, both professionally and personally, and I will be forever grateful for his role in my life.

vi

VITA

June 2001……………Hilliard Darby High School, Hilliard, Ohio

May 2005……………Miami University, Oxford, Ohio Bachelor of Arts in Microbiology

Sept. 2005-present...... The Ohio State University, Columbus, Ohio Doctor of Philosophy in Molecular, Cellular and Developmental Biology

May 2009 ...... Oral presentation, “Developing RNAi therapy for FSHD.” American Society of Gene Therapy, 12th Annual Meeting, San Diego, CA

May 2009 ...... Travel Award, American Society of Gene Therapy, 12th Annual Meeting, San Diego, CA

November 2009 ...... Oral presentation, “DUX4 promotes FSHD-associated pathology in vivo.” Facioscapulohumeral Muscular Dystrophy 2009 International Research Consortium Meeting, Watertown, MA

January 2010-2011 .....Nationwide Children’s Hospital Outstanding Graduate Student Award for 2010, accompanied by stipend fellowship for 2010. The Research Institute at Nationwide Children’s Hospital, Columbus, Ohio

April 2010 ...... Poster presentation, “DUX4, a facioscapulohumeral muscular dystrophy (FSHD) candidate gene, causes muscle toxicity in vivo.” 9th Annual OSU Medical Center Graduate and Postgraduate Research Day, Columbus, OH.

April 2010 ...... Poster presentation winner accompanied by a Travel Award, 9th Annual OSU Medical Center Graduate and Postgraduate Research Day, Columbus, OH.

July 2010 ...... Oral presentation, “Developing RNAi therapy for FSHD candidate genes.” Interdisciplinary Graduate Programs in the Molecular Life Sciences Symposium, Columbus, OH

vii

July 2010 ...... Best parallel session talk award, Interdisciplinary Graduate Programs in the Molecular Life Sciences Symposium, Columbus, OH

November 2010 ...... Travel Award, 60th Annual American Society of Human Genetics meeting, Washington, D.C.

November 2010 ...... Oral presentation, “DUX4 over-expression recapitulates FSHD- associated phenotypes in vivo.” 60th Annual American Society of Human Genetics meeting, Washington, D.C.

November 2010 ...... Outstanding Pre-Doctoral Basic Award, 60th Annual American Society of Human Genetics meeting, Washington, D.C.

November 2010 ...... Poster presentation, “Developing RNAi therapy for FSHD candidate genes.” 60th Annual American Society of Human Genetics meeting, Washington, D.C.

November 2010 ...... Oral presentation, “Targeting FRG1 as proof‐of‐principle for RNAi therapy of dominant muscular dystrophies” Facioscapulohumeral Muscular Dystrophy 2010 International Research Consortium Meeting, Watertown, MA

January 2011 ...... TRINCH 2010 Outstanding Graduate Student Travel Award, Columbus, OH

May 2011 ...... Travel Award, American Society of Gene Therapy, 14th Annual American Society of Gene and Cell Therapy Meeting, Seattle, WA

May 2011 ...... Oral presentation, “RNA Interference improves myopathic phenotypes in mice over-expressing FRG1.” 14th Annual American Society of Gene and Cell Therapy Meeting, Seattle, WA

Aug. 2011-present ...... OSU/Nationwide Children’s Hospital Muscle Group Fellowship, The Ohio State University

October 2011 ...... Oral presentation, “RNA interference inhibits DUX4-induced muscle toxicity in vivo: Implications for a targeted FSHD therapy.” 16th International Congress of the World Muscle Society, Almancil, Algarve, Portugal

October 2011 ...... Oral presentation, “RNAi therapy for dominant LGMD1A.” 16th International Congress of the World Muscle Society, Almancil, Algarve, Portugal

viii

November 2011 ...... Oral presentation, “RNA interference inhibits DUX4-induced muscle toxicity in vivo: Implications for a targeted FSHD therapy.” Facioscapulohumeral Muscular Dystrophy 2011 International Research Consortium Meeting, Watertown, MA

March 2012 ...... Poster presentation, “The DUX4 promoter is preferentially expressed in FSHD-affected tissues.” 2012 Muscular Dystrophy Association Clinical Conference, Las Vegas, NV

March 2012 ...... Outstanding Abstract, 2012 Muscular Dystrophy Association Clinical Conference, Las Vegas, NV

May 2012 ...... Travel Award, 15th Annual American Society of Gene and Cell Therapy Meeting, Philadelphia, PA

PUBLICATIONS

L.M. Wallace, J. Liu, J.S. Domire, S.E. Garwick-Coppens, S.M. Guckes, J.R. Mendell, K.M. Flanigan, and S.Q. Harper. (2012) RNA interference inhibits DUX4-induced muscle toxicity in vivo: Implications for a targeted FSHD therapy. Molecular Therapy Epub Apr 17, 2012.

L.M. Wallace, S.E. Garwick-Coppens, R. Tupler, and S.Q. Harper. (2011) RNA interference improves myopathic phenotypes in mice over-expressing FSHD Region Gene 1 (FRG1). Molecular Therapy Nov;19(11):2048-54.

R.L. Boudreau, S.E. Garwick-Coppens, L.M. Wallace, J. Liu, and S.Q. Harper. (2011) “Rapid Cloning and Validation of MicroRNA Shuttle Vectors: A Practical Guide.” RNA Interference Methods. Ed. S.Q. Harper. Humana Springer Press, 2011, pages 19- 37.

L.M. Wallace, S.E. Garwick, W. Mei, A. Belayew, F. Coppee, K.J. Ladner, D. Guttridge, J. Yang, and S.Q. Harper. DUX4, a candidate gene for Facioscapulohumeral muscular dystrophy, causes p53-dependent myopathy in vivo. (2010) Annals of Neurology, Epub Oct 28; March;69(3):540-52, 2011.

Z. Jin, L.M. Wallace, S.Q. Harper, and J. Yang. PP2A:B56, a substrate of caspase-3, regulates p53-dependent and p53-independent apoptosis during development. (2010) Journal of Biological Chemistry, 285(45):34493-502.

ix L.M. Wallace, S.E. Garwick, and S.Q. Harper. (2010) “RNAi therapy for dominant muscular dystrophies and other myopathies.” Muscle Gene Therapy. Ed. D. Duan. Springer Press. Pages 99-116.

FIELDS OF STUDY

Major Field: Molecular, Cellular, and Developmental Biology

x

TABLE OF CONTENTS

P a g e

ABSTRACT ...... ii

ACKNOWLEDGMENTS ...... v

VITA ...... vii

LIST OF TABLES ...... xiv

LIST OF FIGURES ...... xv

CHAPTERS

1. INTRODUCTION ...... 1

1.1 Introduction ...... 1 1.2 Historical Perspective of FSHD ...... 2 1.3 Presentation and Detailed Clinical Features ...... 4 1.4 FSHD Genetics ...... 7 1.5 FSHD Candidate Genes ...... 9 1.5.1 FRG1 ...... 10 1.5.2 DUX4 ...... 12 1.6 Current FSHD Therapy ...... 14 1.7 Summary and Significance ...... 17

2. RNAi THERAPY FOR DOMINANT MUSCULAR DYSTROPHIES AND OTHER MYOPATHIES ...... 20

2.1 Introduction ...... 20 2.2 Prevalence of Dominant Myopathies ...... 20 2.3 RNA Interferance ...... 21 2.4 RNAi Pathway ...... 22 2.5 RNAi Therapeutics ...... 24 2.6 Disease Allele-Specific Gene Silencing ...... 27 2.7 RNAi Therapy for the Most Common Dominant Muscular Dystrophies 30 2.8 Summary...... 32

xi

3. DUX4, A CANDIDATE GENE FOR FSHD, CAUSES P53-DEPENDENT MYOPATHY IN VIVO…..……..………………………………………..…….38

3.1 Introduction…………………………………………………………….38 3.2 Materials and methods…………………………………….……...……41 3.3 Results………………………………………………………………….46 3.4 Discussion……………………………………………………………...53

4. RNA INTERFERENCE IMPROVES MYOPATHIC PHENOTYPES IN MICE OVER-EXPRESSING FSHD REGION GENE 1 (FRG1)…………...77

4.1 Introduction…………………………………………………………….77 4.2 Materials and methods…………………………………….………...... 79 4.3 Results………………………………………………………………….82 4.4 Discussion……………………………………………………………...86

5. RNA INTERFERENCE INHIBITS DUX4-INDUCED MUSCLE TOXICITY IN VIVO: IMPLICATIONS FOR A TARGETED FSHD THERAPY ...... 106

5.1 Introduction ...... 106 5.2 Materials and methods ...... 107 5.3 Results ...... 111 5.4 Discussion ...... 114

6. GENERAL DISCUSSION ...... 129

6.1 Relevance of this work ...... 129 6.2 DUX4 discussion ...... 129 6.3 RNAi therapy discussion ...... 132 6.4 Concluding remarks ...... 135

APPENDICES

A. RAPID CLONING AND VALIDATION OF MicroRNA SHUTTLE VECTORS: A PRACTICAL GUIDE ...... 137

A.1 Introducton ...... 137 A.2 MiRNA biogenesis ...... 137 A.3 Artificial Inhibitory RNAs ...... 139 A.4 Advantages of miRNA shuttles ...... 141 A.5 Design of miRNA shuttles...... 143 A.6 Primer Design and Cloning of miRNA Shuttles ...... 148 A.7 Rapid Screen to Identify Functional MiRNA Shuttles ...... 152 A.8 Summary ...... 155

xii B. GFP A WORD OF WARNING ...... 166

B.1 Introduction ...... 166 B.2 Materials and Methods ...... 167 B.3 Results and discussion ...... 168

LIST OF REFERENCES ...... 176

xiii

LIST OF TABLES

Table Page

2.1 Mutations causing dominant muscular dystrophies and other myopathies ...... 33

3.1 Total apoptotic gene expression changes in DUX4-injected muscles ...... 60

3.2 Significantly Changed p53 Pathway Genes in DUX4-Injected Muscles...... 62

A.1 Example transfection worksheet for miRNA efficacy screen using pSICHECK-

GOI ...... 157

B.1 Endotoxin report ...... 171

xiv

LIST OF FIGURES

Figure Page

2.1 MicroRNA biogenesis pathway ...... 35

2.2 Natural microRNA sequences and structures are used to design therapeutic inhibitory RNAs ...... 36

2.3 Hypothetical example of mutant allele-specific Cav-3 targeting ...... 37

3.1 V5 epitope tagged DUX4 caused apoptosis in vitro ...... 63

3.2 DUX4 over-expression is detrimental to developing zebrafish muscle ...... 64

3.3 DUX4 toxicity requires DNA binding ...... 66

3.4 DUX4 is toxic to adult mouse muscle in vivo ...... 68

3.5 In vivo DUX4 mRNA and protein expression ...... 70

3.6 DUX4-transduced myofibers are TUNEl-positive ...... 72

3.7 DUX4-transduced myofibers are Caspase-3 positive ...... 73

3.8 DUX4 causes apoptosis through a p53-dependent mechanism ...... 74

3.9 DUX4-transduced myofibers express cleaved Caspase-3 ...... 76

4.1 miFRG1 sequences ...... 89

4.2 In vitro screen to identify lead miFRG1 sequences ...... 93

4.3 FRG1 gene silencing improves muscle mass in FRG1-high mice ...... 95

4.4 AAV.miGFP transduction in 3 week-old FRG1-high mice intramuscularly injected with 5X1010 DRP AAV.miGFP at post-natal day 1 (P1) ...... 97

4.5 HrGFP transduction in isolated muscles from male mice used in this study ...... 98

xv

4.6 In vivo knockdown of FRG1 in FRG1-high mice ...... 101

4.7 FRG1 gene silencing improved myopathic histology in FRG1-high mice ...... 102

4.8 FRG1 gene silencing improved strength in FRG1-high mice ...... 104

4.9 Additional controls for the grip strength assay (Figure 4.8) ...... 105

5.1 DUX4 microRNAs ...... 120

5.2 In vitro screen of miDUX4 sequences ...... 121

5.3 DUX4-targeted artificial miRNAs improve DUX4-associated myopathy ...... 123

5.4 DUX4 knockdown in vivo ...... 125

5.5 miDUX4-treated myofibers are caspase-3 negative ...... 126

5.6 AAV.miDUX4 protects mice from DUX4-associated grip strength deficits ...... 128

A.1 MiRNA biogenesis pathway ...... 158

A.2 Artificial miRNA shuttles derived from natural human mir-30a ...... 160

A.3 Designing and cloning miRNA shuttle vectors: Steps 1-4 ...... 161

A.4 Desiging and cloning miRNA shuttle vectors: Steps 5-11 ...... 162

A.5 Generating a luciferase reporter plasmid with the Gateway® system ...... 163

A.6 Overview of Luciferase assay screen for testing miRNA gene silencing ...... 165

B.1 HrGFP toxicity is dependent on dose and time ...... 172

B.2 EGFP is better tolerated than hrGFP in muscle ...... 174

xvi

CHAPTER 1

INTRODUCTION

1.1 Introduction

Today we know Facioscapulohumeral muscular dystrophy (FSHD) is an autosomal dominant disorder and the most common muscular dystrophy1. These facts were among the few we knew about FSHD for decades. As the name suggests, the muscles of the face and scapula are the most commonly affected muscles, however this distinction took over a century to discern. From the time primary muscle diseases were first separated from anterior horn cell disease in 1868, there have been compounding complications in defining FSHD mostly due to the extreme variability in the myopathy.

This includes severity, age of onset, degree of other muscular and non-muscular involvement as well as extremely complicated genetics. What was started in the late

1800s is still being worked on today. In the last 20 years there has been a major focus on defining the FSHD molecular mechanism, which has opened the door for new FSHD research including this work; our contribution to understanding the genetics of FSHD and developing possible treatment strategies for this very intriguing disease.

1 1.2 Historical Perspective of FSHD

Two French physicians, Louis Landouzy and Joseph Dejerine, first described

FSHD in the late 1800’s2. Over an 11-year period they followed an afflicted family and

documented all the main elements of FSHD. This included observations of progressive

wasting and weakness of the face, scapular and humeral muscles. This was of particular

importance because during the time of their studies the general consensus form the

medical field was that all primary muscular diseases were different manifestations of a

single disease with essentially the same histological changes and intermediate forms of

clinical syndromes2. The observation of primary facial muscle involvement was the first

sign that FSHD was indeed its own disease entity and subsequently spurred years of

debates. Along with their thorough description of the affected muscles, Landouzy and

Dejerine also described the hereditary nature of FSHD and the clinical variability among

affected members of the same family. These observations were strengthened in 1921

when Weitz recognized the possibility of various modes of inheritance (autosomal

dominant, autosomal recessive and x-linked recessive) for the myopathies2. The

introduction of genetic criteria would later prove to be invaluable in solving many FSHD

related disputes.

The next major natural history documentation came in 1950 when Tyler and

Stephens described a family with FSHD members in the United States3. This Utah

family included 1,249 individuals who were all descendants of one affected man. They

identified 58 affected members of the 240 studied. This study confirmed an autosomal

dominant inheritance pattern as well as the clinical heterogeneity3. However, this was still a controversial point, as there seemed to be two schools of thought at this time. One

2 school did not believe facial weakness was a sufficient criterion for separation into its own disease. This group believed a mixed inheritance pattern. Contrarily, Tyler,

Stephens and others believed in the significance of the facial muscle classification and in turn an autosomal dominant inheritance pattern that followed this classification2,3.

Subsequent to the Utah kindred study was the hallmark thesis project of George Padberg in 1982, which was a retrospective review of 19 Dutch kindreds, including 107 affected patients. Here, Padberg provided tremendous details of presenting complaints, clinical progression and laboratory findings in patients with FSHD2. Following his thesis work,

Padberg went on to lead an international consortium in 1991 to establish the major clinical criteria for diagnosis of FSHD4. They included:

1) Onset of the disease is in facial or shoulder girdle muscles

2) Facial weakness affecting eye closure and the mouth occurs in the majority of

patients though weakness may be very discrete and asymmetric

3) Scapular fixators are the most prominently involved muscles in the shoulders

while the deltoids are the least

4) Asymmetry of involvement and weakness is the rule not the exception

5) Progression is inevitable and non-reversible though highly variable and in some

cases, hardly noticeable

6) Hearing loss and retinal disease are part of FSHD but cardiac involvement is not

(though cardiac involvement has more recently been included5-8)

7) Mental retardation can be involved though should not be used for diagnosis

purposes

3 The goals of the consortium and these criteria were multifaceted. The main purpose was

to define and establish uniform criteria not only for diagnosis and identification of

patients, but also for inclusion standards for linkage analysis studies focused on

identifying the FSHD gene. Major advances in FSHD genetics have provided molecular

diagnostic tests and allowed for the expansion and inclusion of other clinical phenotypes,

but the “FSHD gene” has remained one of the most elusive aspects of FSHD research.

1.3 Presentation and Detailed Clinical Features

The majority of FSHD patients have clinical presentations by the second or third

decade of life1. With the exception of affected parents carefully watching for symptoms

in their young children, cases tend to present when patients begin to have difficulty

performing activities over their heads. Because of the everyday life functional aspect

connected with shoulder girdle movements including drying hair, shaving or placing

things on shelves, this weakness is often reported as the first symptom2. While this is

common, it is not all-inclusive as other patients first complain of pain, facial weakness,

foot drop or even pelvic girdle weakness9. Whatever the first symptom may be, there are

a few other generalizations that can be made for the majority of FSHD. The disease is

generally progressive, asymmetric, and tends to move in a descending fashion through

the body2,10,11. Because of the latter, the following clinical features will also be presented in descending fashion.

Facial muscle weakness is the most common feature of FSHD and occurs in 94%

of patients2. Classically facial muscles were defined as those innervated by the seventh cranial nerve. Of these, the most commonly affected muscles are orbicularis oculi

4 (around the eye) and the orbicularis oris (around the mouth). Weakness around the eye

can lead to the inability to completely close eyes, which is the major health concern

connected with the facial muscle weakness in FSHD9. Though originally considered a

quirk, this can lead to serious eye problems like corneal scarring and exposure keratitis.

Another symptom of facial muscle weakness includes swallowing difficulties. A small

percentage of patients present with tongue weakness though the lack of strength in the

orbicularis oris has been linked to most swallowing problems12. The major concern for

FSHD patients, however, is not simply a physical weakness but a social weakness. The

inability to whistle, drink through a straw or blow bubbles may not seem that physically

devastating, but these particular muscles of the face controlling those actions also leave the patients expressionless. The inability to smile or communicate using facial expression is lost leaving patients extremely self-conscious and socially disabled by this

FSHD specific characteristic9.

As mentioned previously, shoulder girdle weakness is often the first noticeable

symptom and is indeed extremely common in FSHD occurring in 93% of patients2.

Various muscle groups are included in this classification including the pectoralis,

sternocleidomastoid, rhomboids, serratus anterior and muscles of the lower and middle

trapezius; which stabilize the scapula2,9. When this weakness occurs, the scapula is destabilized and the clinical feature known as scapular winging occurs. This is one

feature that distinguishes FSHD from other muscular dystrophies. Scapular weakness

leaves patients without the ability to raise their arms above their shoulder9,10. The

humeral aspect (upper arm weakness) typically develops later in the disease, though this

too can be quite variable. The biceps and triceps atrophy quickly while the deltoid and

5 lower arm muscles remain mostly unaffected2. Complimentary to all symptoms described in FSHD, though, there seems to be exceptions to every generalization.

The trunk and lower-extremities can also be affected in FSHD. This includes but is not limited to: abdominals, pelvic girdle, ankle and foot dorsiflexor and leg muscles.

Often this weakness occurs with disease progression, but approximately 5-12% percent of

cases have weakness in one of these muscles as the first presenting symptom9. This can

be a particularly confusing diagnosis in the case of the pelvic girdle due to the similarities

to the limb girdle muscular dystrophies. Lower abdominal weakness can be quite prominent early in disease progression and can result in pelvic tilt, a protruding abdomen and a swayed back posture known as lumbar lordosis2. Ankle dorsiflexor weakness

occurs in 67% of patients and can result in foot drop, a condition that makes walking and

climbing stairs difficult, putting patients at a very high risk of tripping over their own feet

and falling2,9. Advancements in magnetic resonance imaging (MRI) have also indicated

the hamstring, tibialis anterior and gastrocnemius muscles can all be affected13,14.

Considering this kind of muscle involvement it is not surprising that 15-20% of patients

eventually require wheelchair support1.

Global respiratory muscle and cardiac involvement are rarely discussed in FSHD because they have only recently been added to the list of affected muscles5-8,15. With

more advanced testing equipment, clinicians are able to detect more subtle changes in

these muscles. Though this weakness is not as pronounced as in other types of muscular

dystrophy, or even as pronounced as other muscle groups in FSHD, it is still important to

screen patients for this involvement and to treat these symptoms preventively.

6 Finally, non-muscular features include retinal abnormalities, hearing loss and

central nervous system defects. Retinal vasculopathy consisting of capillary

telangiectasia, aneurysms and capillary closure occurs in 60% of patients, though only a

very small percentage loses sight16. On the other hand, up to 75% of patients suffer from

hearing loss16. Like other FSHD symptoms, hearing loss can be progressive, starting

with high tone perceptive deafness and progressing to include all frequencies. While the

latter two symptoms affect the general FSHD population, central nervous system deficits

including mental retardation and epilepsy have only been shown in a few severely

affected children17,18.

1.4 FSHD Genetics

Clinical features were the key factor in identifying FSHD as a unique disease, but

the genetics were the key to proving it. The first major genetic breakthrough for FSHD

came in 1990 when linkage analysis from 10 Dutch families showed the FSHD locus was

mapped to the subtelomeric region of chromosome 419. This region of 4q35 was shown

to contain 3.3 kilobase repeat elements that were subsequently designated D4Z4 repeats.

In 1992, it was found that FSHD was caused by a contraction of these repetitive elements

leaving patients with one allele containing a contracted macrosatellite repeat array of 1-

10 D4Z4 units compared to the general population which contained 11-150 units20. The next logical step was to find out what other features were in the area of the array, but this proved to be more difficult than anticipated. Following the discovery on chromosome 4, it was established that another D4Z4 repeat array with 99% sequence identity was located on chromosome 10. Both 4q35 and 10q26 D4Z4 arrays were shown to be highly

7 polymorphic and exhibit extensive size variation21,22. Indeed contractions on 10q26 were also found, but they were associated with 10% of the normal population and not with

FSHD23. FSHD patients with a double contraction on both chromosomes were also identified and studied to see if 10q26 contractions were acting as a modifier to the disease, but nothing was found to support that hypothesis24.

In 1996, an individual was identified that was monosomic for the region, meaning complete loss of 4q35 D4Z4 repeats on one allele. Surprisingly, this deletion was not associated with FSHD25. This unique mutation indicated that a minimum of one chromosome 4 D4Z4 repeat was necessary for disease manifestation, but at the time no known genes existed within the actual repeat sequence. Each D4Z4 unit does contain a copy of the DUX4 gene (double homeobox 4), though this was initially discarded as

“junk DNA”26-28. Because no genes were lost or mutated in the contracted region, FSHD did not seem to follow classic Mendelian genetics. Instead the focus of FSHD genetics was shifted to D4Z4 associated genes, other disease causing regulatory sequences potentially located nearby, and intense sequence analysis for allele specific polymorphism identification.

The new problem became how to distinguish between the extremely homologous type 4 and type 10 contractions since only the type 4 was associated with the disease.

Luckily slight sequence differences were identified between the 4q35 array and the 10q26 array, and a molecular diagnostic test was created that could reliably the two chromosomal arrays29. While this test is still used today, it cannot account for all of the complexities of FSHD genetics. A few more compounding factors added to the diagnostic nightmare of FSHD, but also, ultimately led to the current genetic mechanism.

8 First, it was found that the 4q35 and 10q26 arrays were so similar because the 10q26

formed following a duplication and translocation of the 4q35 array while only gaining

minute sequence variations in the process30. As mentioned previously, this led to hybrid

D4Z4 arrays containing both 4 and 10 types that were shown to exist in 10% of people,

both normal and FSHD associated. This indicated that there was part of the FSHD

population with contracted hybrid alleles. Finally, this complexity reached a pinnacle

when different 4qter variants were identified. These differed only by the presence of a

6.2 kilobase b-satellite sequence. These variants, designated 4qA (with b-satellite

sequence) and 4qB (without b-satellite sequence), are found equally distributed in the

population but FSHD is only associated with the 4qA variant30. Eight years following the

variant discovery, a more rare variant designated 4qC was identified, but like 4qB, it was

also found to be non-permissive for FSHD31. In the last few years additional sequences

were identified which further subdivided the variant designations32,33. Additional

polymorphisms subdivided both chromosomes into at least 17 genetically distinct

variants for chromosome 4, and 8 variants for chromosome 10. Of these only three

chromosome 4 variants were associated with disease meaning FSHD manifestation is completely dependent on a few very specific permissive haplotypes. The reason for this will be discussed later.

1.5 FSHD Candidate Genes

There is abundant controversy in the literature surrounding the FSHD gene. Lack of a detectable transcript from the D4Z4 repeats coupled with epigenetic regulation in the array area served as the basis for the current FSHD pathogenic mechanism. In the last 10

9 years, the search for the causative gene has turned up many FSHD gene possibilities, all

reliant on the same mechanism stating: a contraction of D4Z4 repeats results in an altered

chromosomal structure leading to position effect variegation. Indeed D4Z4 repeats are

normally embedded in chromatin and therefore transcriptionally silent, however, in

FSHD the D4Z4 repeats are hypomethylated leading to an open chromatin structure that

is more permissive for gene expression. Other epigenetic factors also support this model

including the discovery of weakened nuclear matrix attachment sites, and a near by

transcriptional enhancer element34-37. Thus, FSHD results when the D4Z4 contraction

induces disassociation from the heterochromatin, and therefore a derepressive effect on

D4Z4 associated genes leading to up-regulation or over-expression of associated

causative gene(s). In this light, the FSHD region (defined as several hundred kb

upstream of the array) was searched in hopes of identifying possible genes.

1.5.1 FRG1

The first possible FSHD candidate gene was found in 1996 and bore the name of

the region, FSHD region gene 1 (FRG1)38. It was located 100kb upstream of the D4Z4

array and in time would be considered to play a prominent role in the disease. The function of FRG1 is still being investigated, but to date, it has been shown to be involved in muscle development, angiogenesis, alternative RNA splicing, actin bundling and muscle cell proliferation39-45. Other genes were also discovered in the FSHD region

including ANT1 and FRG238,39,46. In 2002, a small number of FSHD muscle samples were screened for these genes and all were shown to have increased expression levels compared to controls35. These three genes also seemed to fit the epigenetic FSHD model

10 because the most proximal gene to the D4Z4 array, FRG2, was also the most highly up-

regulated, thus demonstrating a distance-dependent effect35. As a follow-up to the gene expression study the same lab created multiple lines of transgenic mice expressing human

ANT1, FRG1 or FRG2 from the human skeletal α-actin (HSA) promoter. Of all the mice created, only those over-expressing FRG1 showed a myopathic phenotype, which

importantly resembled phenotypes seen in FSHD patients39. Three independent lines

were created and named FRG1-high, -medium and -low respective to their expression

level. Logically the FRG1-high mouse was the highest expressing (40-fold over-

expressing) and demonstrated the most severe phenotypes. All lines of FRG1 mice

showed a progressive deterioration much like the progressive nature of FSHD.

Phenotypes were very distinct and confined to the skeletal muscle, though the latter is

due to the specificity of the promoter.

Phenotypes of the FRG1 mouse can be detected from simple gross anatomy to

molecular analysis. Just by visually looking at the mouse differences can be seen

compared to controls. FRG1 mice tend to be smaller in stature, which is detectable

virtually from the time of birth in the case of the FRG1-high mouse. As all the lines age,

the mice develop mild to severe kyphosis (outward curvature of the spine), which is the

rodent equivalent of the human condition lordosis. This condition was shown to be a

result of weakened muscles of the trunk and not due to skeletal deformities39. In addition

to reduced body weight, muscle size is also markedly reduced especially in the limbs.

Not all muscle groups are atrophied equally however, which is one distinct similarity to

FSHD patients. In addition, FRG1 over-expressing muscles show dystrophic histological

features including variable fiber size, necrosis, central nuclei, and fibrosis. Not

11 surprisingly, these muscles also show functional deficits by way of a reduced exercise

tolerance that can be assessed by various behavioral tests such as treadmill training or

grip strength. The original description of the FRG1 mice also included some molecular

analysis of pre-mRNA splicing39. Multiple genes affected in myotonic dystrophy,

another myopathy caused by misregulated alternative splicing, were analyzed and

compared to muscle cell cultures derived from FSHD patients. All genes were found to

have similar expression profiles39. At the time, these mice were considered a good animal model for FSHD and consequently the best resource for studying the molecular basis of pathogenesis and evaluating potential therapeutics. However, later studies found

FRG1 was not uniformly increased in all patient biopsies and though the mice develop a consistent dystrophy, they were not sufficient evidence to rule out the possibility of other genes being involved in FSHD pathogenesis47,48. Today the role of FRG1 in FSHD is

controversial and the field generally considers FRG1 to be at most a minimal contributor

to FSHD pathogenesis.

1.5.2 DUX4

DUX4 is currently the leading FSHD candidate gene. Until 2010, however, the

role of DUX4 in FSHD was strongly debated and controversial. DUX4 was first

identified about 17 years ago28. Its location within each D4Z4 repeat would have made it the obvious FSHD candidate gene, but at the time there was no obvious promoter, polyadenylation (polyA) signal or any introns associated with the gene21,28,49,50. Added to

the fact that the transcript was completely undetectable, DUX4 was understandably

designated a non-functional retro-transposed pseudogene. Unbeknownst to FSHD

12 researchers, many of these elements were being missed due to the technical difficulties

associated with the region. First, DUX4 transcripts seem to be expressed in very low

abundance. One study indicates only 1 in 1,000 FSHD muscle cell nuclei express the

gene51. Second, the region is incredibly GC rich making sequencing particularly difficult. Lastly the polyA is only present beyond the distal repeat in a flanking region designated the pLAM sequence32,52,53. Only permissive FSHD alleles contain a pLAM sequence that provides a stabilizing polyA for DUX432. This was confirmed with the

identification of a FSHD family that presented with a contracted hybrid allele containing

a translocation of the chromosome 4 distal region (including the pLAM sequence) onto

chromosome 10. This evidence confirmed the importance of the DUX4 polyA and excluded all other chromosome 4 genes from being the primary FSHD insult. Prior to

2010, however, it took a concentrated effort from a small group of researchers to bring

DUX4 back into the FSHD picture.

DUX4 re-emerged as an FSHD candidate gene in 2007. At this time, three

independent groups published intriguing data on DUX4. The first group showed

evolutionary data suggesting the DUX4 gene was conserved by selection. Specifically,

the DUX4 open reading frame (ORF) had been conserved for over 100 million years.

Indeed non-human primate and human D4Z4 sequences have virtually no sequence

similarity outside the DUX4 ORF54. This data was the first to refute the DUX4

pseudogene theory. Second, DUX4 over-expression was shown to cause apoptosis in

myoblast cultures53. This data made DUX4 the first candidate gene to display overt cell

toxicity. Finally, and arguably most importantly, the third group identified DUX4 as a

transcription factor with the potential to regulate many genes. In particular, they showed

13 DUX4 binds to the promoter and up-regulates paired-like homeodomain transcription

factor 1 (PITX1)52. This gene is of particular interest in FSHD because it is up regulated

in patients and functions to control hind limb development as well as regulate right/left symmetry. As a part of the same study, this group also demonstrated that polyadenylated

DUX4 mRNAs were detectable in immortalized FSHD myoblast cell lines via RT-PCR52.

Together these studies defined DUX4 not only as an expressed transcript, but also as a

coded transcription factor that could potentially cause severe damage to muscle as well as

contribute to the hallmark characteristic of asymmetry seen in FSHD patients. At this

time we began our research on DUX4 exploring the toxic potential of the gene when

expressed in vivo. This work will be discussed in detail in Chapter 3.

Post 2007, our work and others have focused on defining the pathophysiological

potential of DUX4. Though the normal biological function is largely undefined, we are

beginning to understand its potential role in FSHD. Indeed DUX4 seems to play a

causative role in muscle differentiation defects due to interference with myogenic

regulatory factors, and expression of the gene is extremely toxic to a plethora of cell

types53,55,56. Its role as a transcription factor could impact a large number of genes, which

may explain certain discrepancies seen between various gene expression studies

published to date. Coupled with the evidence of the stabilizing poly A for DUX4 only

being present on FSHD-permissive chromosome variants, DUX4 is likely the primary

causative gene in FSHD and subsequently the preeminent target for a FSHD specific

therapy.

1.6 Current FSHD Therapy

14 Currently, there is no FSHD-specific treatment. Because the pathogenic

mechanism was unresolved and is still being fully worked out, rational disease-specific therapies (genetic or pharmaceutical) have understandably not been developed. Instead, general muscular dystrophy drugs have been considered as well as exercise therapy in addition to assistive devices and specific surgical procedures.

One treatment strategy has been developing pharmaceuticals to increase muscle mass and strength. There have been clinical trials testing the efficacy of three different drugs to date: prednisone, albuterol and MYO-02957-61. Prednisone was the first clinical

trial for FSHD. It was considered a very promising drug due to the positive responses

seen in Duchenne muscular dystrophy and Becker muscular dystrophy patients.

Prednisone is a corticosteroid that significantly delays the symptoms of other muscular

dystrophies, but not in FSHD58. Albuterol was the second clinical trial. The drug is a

beta(2)-adrenergic agonist that prevents atrophy by exerting anabolic effects on the

muscle. While the pilot trial showed promise, the randomized, double-blind, placebo-

controlled study failed to show significant improvement in global strength or

function57,60. Though albuterol did increase muscle mass and some measures of strength

in patients, it was not adopted as a standard of care for patients due to the non-significant

functional data. Along the same line as the albuterol trial goal (increasing muscle mass),

most recently, the general buzz in the muscular dystrophy field has been the manipulation

of myostatin. Myostatin is part of the transforming growth factor-β superfamily and is an endogenous inhibitor of muscle growth62. Indeed, disruption of the myostatin signaling

pathway has been shown to significantly enlarge muscle mass in multiple animals63.

MYO-029 is a neutralizing antibody to myostatin that was used in the most recent FSHD

15 clinical trial. Unfortunately, a cutaneous hypersensitivity reaction precluded dose

escalation, and though low doses were deemed “safe”, they also did not show any

significant improvement in strength or function59. To date, no general pharmaceutical

has shown efficacy against FSHD.

All other treatments for FSHD have been directed at specific symptom

management. Ankle muscle weakness and scapular muscle weakness have both been

treated using orthoses. For the ankle, braces most commonly recommended include

fixed or hinged ankle-foot orthoses or floor reaction orthoses9. These types of devices

aid with difficulties associated with foot drop. The shoulder girdle can also benefit from

use of specific orthoses. Though it is not a common use, one report described benefits of

a spinal orthosis to aid with scapular weakness64. More frequently however, scapular fixation is used65. A stabilized scapula aids with overhead activities, pain associated with

scapular winging, as well as correcting the physical deformity. Taping and slings have

been used, but these can limit mobility. Other options include a scapular fixation

surgery, like a scapulodesis, where the scapula is fixed to the thoracic wall with screws,

wires, or plates to produce a solid fusion65. Obviously all these options have risks and

benefits to the patients so each case must be individually assessed and carefully

considered before treatment is prescribed. Furthermore, as with many modes of disease

management, practices and timing of interventions can vary greatly from clinician to clinician making clinical trials difficult if they are reliant on similar patient treatment

backgrounds.

One trend in multiple clinical trials, including one that is currently underway,

looks at the role of physical training or exercise61. The benefits of exercise in the normal

16 population are well defined. However, exercise is controversial in the muscular

dystrophy population. The concern is based on the idea that overuse could actually

exacerbate and perpetuate weakness. Few trials have focused on FSHD and all exercise

trials to date have lacked in standardization, and power. Currently the overall basic

recommendation for FSHD patients is to pursue an active lifestyle that includes flexibility

training and moderate-intensity aerobic and resistive strength training9.

1.7 Summary and significance

FSHD can be a debilitating disease emotionally, socially, and physically. The

exact incidence can vary between regions, but is estimated to be around 1 in 7,5001. To

date, all pharmaceutical trials for FSHD have failed and there is no consensus on what the

standard of care should be in relation to symptom management. FSHD patients are in

desperate need of a treatment specifically targeted to their disease. Chapter 2 is an

introduction to RNA interference (RNAi) and its possible role as a tool for treating FSHD

and other dominant muscular dystrophies. Here we discuss the benefits of specifically

targeting disease genes, however, because the etiology for FSHD has been largely

unknown, therapy development including RNAi has suffered. The goals of the studies

presented here are to gain an understanding of the pathogenic disease mechanisms and to

develop targeted translational therapies to treat FSHD.

Because the disease gene for FSHD was still controversial, we began our studies

by testing one of the less defined FSHD candidate genes, DUX4. Based on DUX4’s

location in the D4Z4 repeats and its toxic nature in vitro, we hypothesized that DUX4 over-expression would alter wild type muscle in vivo and that this alteration would be

17 directly linked to its activity as a transcription factor. In Chapter 3 we tested the

myopathic potential of DUX4 in zebrafish and wild-type mice, both of which

demonstrated the toxic nature of the gene. By using a DUX4 DNA binding domain

mutant, we also linked this toxicity directly to the actions of DUX4. Finally, we provided

a potential mechanism for this toxicity by showing DUX4-induced damage is p53-

dependent. Indeed Chapter 3 demonstrates the extreme myotoxic potential of DUX4,

which helped to define its role as the causative gene in FSHD.

Next, we tested the efficacy of RNAi as a treatment for a dominantly inherited

muscular dystrophy. For this study, we utilized the mouse model of the other major

FSHD candidate gene, FRG139. Though we had doubts about FRG1’s actual role in

FSHD, the mouse model was perfect for RNAi proof-of-principle experiments.

Specifically, it was an animal model with a well-defined and consistent myopathic

phenotype. It also did not require allele specific silencing. Due to the design of the

mouse, we could target the over-expressed human gene responsible for causing the

myopathy without targeting the endogenous mouse FRG1 gene. As presented in

Chapter 4, we developed FRG1-targeted artificial microRNAs (miRNA) and delivered them to FRG1 over-expressing mice using adeno-associated viral vectors (AAV).

Importantly we found a single intramuscular injection of the therapeutic miRNA could

prevent symptoms of muscular dystrophy in these mice up to 14 weeks, the longest time

tested. Treated muscles were stronger and showed no histological signs of disease.

Chapter 4 shows proof-of-principle that targeted gene therapy using RNAi could be a

plausible treatment for dominant muscular dystrophies including FSHD.

18 Chapter 5 applies the principles demonstrated in the previous chapter to FSHD

specifically. Recent evidence of DUX4’s causative role in FSHD has made the gene the

most promising therapeutic target32. Here we describe a unique pre-clinical study testing

DUX4-targeted miRNAs (miDUX4). Because no stably expressing DUX4 animal model exists, we designed an AAV co-injection strategy to deliver both AAV.DUX4 and

AAV.miDUX4. This model allowed us to simultaneously induce DUX4 over-expression and silence the gene in vivo. Results were extremely promising as the miDUX4 treatment was capable of protecting muscles in vivo. Importantly, miDUX4 was safe and protective even at the physiologically elevated levels of DUX4 expression present in the model. Together this work helps define the role of DUX4 in FSHD pathology and suggests a DUX4-targeted therapy could be effective in treating FSHD patients.

19

CHAPTER 2

RNAi THERAPY FOR DOMINANT MUSCULAR DYSTROPHIES AND OTHER

MYOPATHIES

2.1 Introduction

Over the last ~15 years, muscular dystrophy gene therapy strategies have been primarily aimed at replacing defective or missing genes underlying recessive disorders, such as Duchenne muscular dystrophy (DMD). These gene replacement strategies are typically not indicated for treating dominant diseases; instead, patients bearing dominant mutations would likely benefit from reduction or elimination of the abnormal allele.

Until very recently, there was no feasible mechanism to reduce or eliminate disease genes, and molecular therapy development for dominant muscular dystrophies was largely unexplored. RNA interference (RNAi) has recently emerged as a powerful tool to suppress any gene of interest in a sequence specific manner. As such, RNAi is a leading candidate strategy to silence dominant disease genes, including those involved in muscular dystrophy and related myopathies. Here, we discuss the potential for RNAi- mediated gene therapy of dominant muscular dystrophies and other myopathies.

2.2 Prevalence of Dominant Myopathies 20 Individually, all myopathies are classified as rare disorders by the NIH Office of

Rare Diseases and Orphanet, which respectively define rare diseases as those affecting

less than 200,000 people in the U.S. or 1 in 2000 Europeans 1,66. Among all muscular

dystrophies, X-linked recessive DMD is the most common (1/3,500 newborn males)1,

followed by the dominant disorders myotonic dystrophy type 1 (DM1; 1/8,000)67 and facioscapulohumeral muscular dystrophy (FSHD; 1/15,000 - 20,000)10,68. However, a

recent Orphanet report of disease prevalence in Europe places FSHD first, followed by

DMD and DM (Table 2.1)1. Mutations in at least 29 known genes cause various

dominant muscular dystrophies and other related myopathies, and there are at least 7

other clinically characterized disorders that have yet to be linked to specific genes (Table

2.1). Collectively, these dominant myopathies approach a prevalence that classifies them

as common disorders (~1/2400 to ~1/3200)1. This is significant since similar gene

silencing strategies, with modifications depending on genetic etiology, may be effective

for treating most dominant myopathies. Thus, proof-of-principle demonstration of RNAi

therapeutic efficacy in one myopathy could broadly impact an entire class of disorders.

2.3 RNA interference

RNA interference (RNAi) is a cellular mechanism to control coding gene expression prior to translation 69. RNAi-controlled messenger RNAs are therefore

transcribed but not translated. RNAi is mediated by small (21-25 nucleotide, nt), non- coding microRNAs (miRNAs) and several proteins involved in miRNA processing and gene silencing 70-72. A key feature of RNAi is sequence specificity: miRNAs share

nucleotide sequence homology and base-pair with 3’ untranslated (UTR) regions of

21 cognate mRNAs 73. These base-pairing interactions allow miRNAs to direct cellular

gene silencing machinery to target mRNAs and prevent their translation.

Naturally occurring miRNAs arise as relatively long primary transcripts from

eukaryotic genomes ranging in complexity from single-celled algae to mammals 71,72,74-77.

Over the last several years, a large amount of work has focused on understanding how

miRNAs are expressed and processed to a biologically functional form. An important

consequence of this growing knowledge has been the development of RNAi therapeutics.

Designed RNAi molecules can be engineered to mimic natural miRNAs and

subsequently used to suppress any gene of interest. It is therefore important to

understand the biology underlying natural microRNA biogenesis when developing RNAi

as a therapeutic tool.

2.4 RNAi Pathway

Rationally designed RNAi molecules are based on the structure and, in some

cases, the nucleotide sequence, of natural miRNAs (Figs 2.1-2.2; 78-80. Like other coding

and non-coding transcripts in the cell, primary miRNA (pri-miRNA) precursors vary in

size (can be several kilobases in length) and how they are transcribed: some miRNAs are

RNA polymerase II (pol II) transcripts, others are RNA polymerase III (pol III), and

expression may be tissue-specific 80-83. Transcription of the pri-miRNA is the first step in the miRNA biogenesis pathway. The pri-miRNA is generated as a single-stranded transcript that forms an intramolecular stem-loop structure. Subsequent post- transcriptional processing steps, catalyzed by several evolutionarily conserved proteins, serve to trim the pri-miRNA to a smaller, functional form, and ultimately create a double-

22 stranded miRNA from the single-stranded primary transcript. Simplistically, the pri-

miRNA contains important sequence and structural elements that direct a nuclear

microprocessor complex composed of Drosha and DGCR8 to cleave the RNA at a

specific location 84-88. DGCR8 recognizes and binds an important miRNA structural

feature – a characteristic junction between the miRNA double-stranded stem and flanking

single-stranded sequences 88. DGCR8 binding serves as a ruler to correctly position

Drosha at the base of the miRNA stem, where it then makes a staggered cleavage to

produce a characteristically shorter (~65-70 nucleotide, nt) hairpin pre-miRNA containing a 2 nt 3’ overhang 87,88. The nuclear export factor, Exportin-5 (Exp5) binds this overhang and then shuttles the pre-miRNA to the cytoplasm 89. There, the enzyme

Dicer recognizes the Drosha-generated overhang and catalyzes another staggered

cleavage event ~21 nt away (~ 2 RNA helical turns), which removes the loop from the

hairpin and produces a second 2 nt 3’ overhang at the opposite end 90-92. The final result

is the mature, 21-25 nt duplex miRNA containing 2 nt 3’ overhangs at both ends. This small range in mature miRNA size may be partly accounted for by bulged, looped-out

mismatches in the miRNA stem. Since a Dicer cut is ~21 nts long, stem mistmatches that

do not extend the length of the RNA helices may still be incorporated in the primary

mature guide strand, which may, as a result, be slightly longer in some natural miRNAs.

One strand of the mature miRNA duplex (the antisense “guide” strand) becomes the

RNA component of the RNA Induced Silencing Complex (RISC), which is ultimately responsible for sequence-specific gene silencing. The sense or “passenger” strand of the

miRNA may be degraded or used to program a second RISC complex 93,94. Indeed, some

miRNAs are bi-functional and both strands can direct gene silencing 94. For therapeutic

23 RNAi strategies, it is therefore important to validate that only the intended guide strand is

directing gene silencing, as this will reduce risks of non-specific, “off-target” effects.

The degree of complementarity between the guide strand miRNA and an mRNA determines (1) whether the transcript will be regulated at all by a programmed RISC complex; and (2) if so, which of two gene silencing mechanisms will be induced

(translational inhibition or transcript degradation). In general, incomplete pairing of an inhibitory RNA and a target mRNA will produce gene silencing through translational inhibition. In this case, target mRNA levels do not change. In some instances, as little as

7 nt of homology between the guide strand and a target mRNA (miRNA nts 2-8; called the seed match) may be required to elicit gene silencing effects 95. Base-pairing outside the seed region may serve to stabilize the miRNA-mRNA interaction and help produce a more robust knockdown. In contrast, perfect miRNA-mRNA complementarity across the

~21-25 nt stretch results in mRNA degradation, and thus the pool of target mRNAs in the cell is depleted. The degradation mechanism is associated with more robust gene silencing.

2.5 RNAi Therapeutics

RNAi molecules can be engineered to suppress any gene. Numerous strategies to design inhibitory RNAs have been developed and all share two common features: artificial RNAi molecules are double-stranded and comprised of sequences cognate to an mRNA of interest. Artificial inhibitory RNAs can be designed to mimic mature, pre-, or pri-miRNAs and will thus, upon delivery to cells, enter the miRNA pathway at different points (Fig 2.1). There are three main classes of inhibitory RNAs (Fig 2.2). (1) Small

24 inhibitory RNAs (siRNAs) are in vitro synthesized, dsRNAs that are structurally identical to miRNA duplexes 71. When delivered to cells, all siRNAs bypass the transcription and

nuclear processing steps of the miRNA pathway. Some designed siRNAs are processed

by Dicer 96, while others avoid this step and are immediately available to complex with

RISC proteins after delivery to the cytosol. (2) Short hairpin RNAs (shRNAs) are

structurally similar to stem-loop pre-miRNAs. They are typically designed to contain

~21 nt of paired stem sequence connected by an unpaired loop that is often derived from natural microRNA sequences97. ShRNAs are produced intracellularly, arising as

transcripts from DNA expression cassettes using RNA pol III, and very rarely, pol II

promoters. ShRNAs mimic Drosha-processed miRNAs and thus, following transcription,

are immediately shuttled by Exp-5 to the cytoplasm for Dicer processing and

incorporation into RISC. (3) Artificial miRNA shuttles resemble pri-miRNAs 78,79. Like

shRNAs, miRNA shuttles are transcribed from DNA expression cassettes, but are

amenable to regulation by both pol II and pol III promoters. In this design, miRNA

sequences required to direct Drosha and Dicer processing are maintained, but the natural,

mature, 21-25 nt miRNA sequence is replaced by an inhibitory RNA sequence targeting

the gene of interest. Thus, a natural microRNA is used to deliver an artificial siRNA.

MiRNA shuttle transcripts are produced intracellularly and utilize all processing steps

required for natural miRNA biogenesis.

Each of the three systems described above are capable of eliciting strong RNAi

responses in vitro and in vivo. The key difference between siRNAs and shRNA/miRNAs

is duration of expression. In vitro synthesized siRNAs are transient and long-term

disease-gene suppression requires repeated administration; expressed shRNAs or miRNA

25 shuttles are longer lasting, and if delivered via an appropriate viral vector, may produce

permanent gene silencing effects. Importantly, muscle-directed gene delivery systems

are well-developed, especially those using adeno-associated viral (AAV) vectors, which

have been used extensively in the last few years to deliver shRNA/miRNA to numerous tissues 98-101.

As described above, shRNAs and miRNAs differ in the level of processing

required by endogenous miRNA biogenesis machinery. This differential processing has

direct implications for how each is expressed. Because shRNAs are not Drosha- processed, their 5’ end must be defined by the start of transcription. This is important because Dicer binds the “Drosha-cut” end of the pre-miRNA and makes a defined cut

~21 nt downstream, which ultimately determines the sequence of the mature guide strand molecule (Fig 2.2). As a result, shRNAs must be positioned near a promoter’s transcription start site to ensure proper processing and gene silencing function. This restriction is not necessary for miRNA shuttles because Drosha processing, not transcription, defines the critical 5’ Dicer binding site. As a result, artificial miRNAs can be expressed from any promoter. Moreover, several bi-functional expression vectors have been described, in which a coding gene and intron- or UTR-embedded miRNA arise from the same pol II promoter-driven transcript 102,103. Another difference between

shRNAs and miRNAs is potential for non-specific toxicity; miRNAs may be safer than

shRNAs in vivo 79. ShRNAs were the first generation of plasmid- or vector-expressed artificial inhibitory RNAs used in vivo. Several studies have demonstrated shRNA efficacy for silencing disease genes and improving associated pathologies in, for example, models of neurodegenerative disease and viral infection 98-100,104. However, a

26 few recent studies have raised concerns about shRNA safety. Specifically, uncontrolled,

high-level shRNA expression from constitutively active pol III promoters caused liver

failure and brain striatal loss in mouse models of hepatitis and Huntington’s Disease

(HD), respectively 100,105. This observed toxicity seems to be related to shRNA-induced saturation of endogenous microRNA biogenesis pathways, especially at the level of nuclear export, thereby interfering with natural microRNA function 100. Importantly,

lowering the dose of vector-expressed shRNAs in the liver, or using a less-powerful

microRNA shuttle system in the brain, mitigated these toxic effects 100,105. Both

strategies ultimately led to significant gene silencing without over-expression associated

toxicity. Although not all shRNAs are overtly toxic, and sufficient safety data regarding

long-term artificial miRNA is lacking, miRNA shuttles may be safer than shRNAs

simply because they are more efficiently processed and amenable to expression by tissue-

specific, regulated, or weaker RNA pol II promoters, while shRNAs are dependent upon

strong, constitutively-active pol III promoter expression. Regardless of the system used,

RNAi therapy has shown promise in pre-clinical models of neurodegenerative disease,

viral infection, and cancer, among others 98-101,104,106,107; these studies support its potential

for treating dominant muscular dystrophies and other myopathies.

2.6 Disease Allele-Specific Gene Silencing

Excepting the extremely rare cases of X-linked dominant FHL1 mutations in males 108, patients with dominant disorders possess one mutant and one normal copy of

their specific myopathy-related gene. Since the underlying pathogenic events in these

disorders are dominant gene mutations, simply reducing mutant allele expression may be

27 therapeutic. For example, myotilin (MYOT) mutations cause dominant, progressive

muscle disease clinically classified as LGMD1A, myofibrillar myopathy (MFM), or

spheroid body myopathy (SBM)109-114. To date, 12 distinct point mutations were

associated with dominant myotilinopathies in numerous families 109-114. Myotilin is a Z-

disc protein expressed predominantly in skeletal and cardiac muscle. Mutations cause

myofibrillar aggregation and muscle weakness that is recapitulated in transgenic mice

expressing a dominant human mutation (T57I)110. Importantly, MYOT null mice are

normal; they show no obvious muscle pathology or weakness, Z-disk proteins are unaltered, and animals live a normal lifespan 115. These data suggest that there may be a

compensatory mechanism to counteract MYOT deficiency in mice. Whether MYOT

absence in well-tolerated in humans is unknown. However, because LGMD1A

phenotypes are recapitulated in an available mouse model, and MYOT absence produces

no overt defects in mice, it may be an ideal target to demonstrate proof-of-principle for

RNAi therapy of dominant muscle disorders.

In contrast to myotilinopathies, which may be an exception, most dominant muscular dystrophies may require specific silencing of the dominant allele. Since normal copies of disease genes likely encode essential proteins, normal allele haploinsufficiency may contribute to myopathic phenotypes as well. Loss-of-function contributions to dominant disease can be predicted from knockout mouse models and by examining genetic case studies, in which different mutations in the same gene give rise to dominant and recessive myopathies. For example, Nemaline Myopathy (NM) can arise from autosomal dominant or recessive TPM3 mutations 116-124. Dominant NM patients have one mutant and one normal TPM3 gene copy, while human carriers of recessive alleles

28 and TPM3 +/- mice are normal, and TPM3 -/- animals die as embryos 124,125. These

observations support two conclusions: only one normal TPM3 allele is required to

maintain normal muscle, and gain-of-function TPM3 mutations are most likely the sole pathogenic event in dominant NM forms. Therefore, an RNAi strategy that specifically suppresses mutant TPM3 while leaving the normal allele untouched may improve myopathy in NM patients. Likewise, disease allele-specific RNAi therapies may be important for Caveolin-3-related myopathies, since normal Cav-3 gene dosage impacts muscle disease severity 126-129. Specifically, severe LGMD1C is caused by autosomal

recessive homozygous or dominant negative Cav-3 mutations resulting in complete or

97% Cav-3 loss. In contrast, different mutations resulting in 84% or 50% Cav-3

reductions produced mild hyperCKemia without muscle weakness, or normal

phenotypes, respectively 126-129. In both NM and LGMD1C examples, it would be

advantageous to restrict gene knockdown to the affected allele while leaving the normal

allele unperturbed. Since many dominant myopathies are caused by single point mutations in one allele, the question arises: can inhibitory RNAs be designed to distinguish two transcripts differing by one base pair? In short, the answer is yes. As previously discussed, perfect sequence complementarity between an inhibitory RNA and target mRNA causes message degradation; imperfect base pairing leads to translational inhibition130. However, this rule is not absolute. Complementarity does not ensure

inhibitory RNA efficacy; not all inhibitory RNAs containing perfect homology with a

target mRNA actually cause gene silencing. Conversely, more mismatch does not

necessarily reflect reduced potency; microRNAs can have several mismatches with a

target mRNA and still cause gene silencing, but a single nucleotide difference may be

29 sufficient to prevent silencing altogether 95,131-135. Thus, well-designed inhibitory RNAs can specifically silence disease genes by distinguishing between normal and mutant alleles differing by one nucleotide. Although each allele-discriminating miRNA must be uniquely designed and empirically validated, some general guidelines can be followed.

Specifically, the discriminating nucleotide should be placed centrally within the inhibitory RNA duplex and if sufficient disease allele-specific silencing is not produced, optimal specificity can be achieved by including additional peripheral mismatches in the inhibitory RNA sequence (Fig 2.3).

2.7 RNAi Therapy for the Most Common Dominant Muscular Dystrophies

Myotonic dystrophy type 1 (DM1) and facioscapulohumeral muscular dystrophy

(FSHD) are the top two most common muscular dystrophies and both are dominantly inherited. Therefore, DM1- and FSHD-targeted treatments would potentially have the broadest benefit for patients with dominant muscle disease, making them logical candidates for RNAi therapy development. However, both disorders have complicated and unique etiologies that make them challenging, though not impossible, targets for

RNAi treatment.

Myotonic Dystrophy Type 1. DM1 is caused by CTG trinucleotide repeat expansion in the DMPK 3’ UTR, which causes nuclear retention of this toxic mRNA 136. Patients

develop myotonia leading to skeletal muscle weakness, and cardiac conduction

abnormalities that often cause death in patients. DMPK +/- and -/- knockout mice both show skeletal and cardiac muscle sodium channel gating abnormalities that recapitulate conduction defects in human DM1 137-139. Older (7-11 month) DMPK +/- mice also show

30 mild, variable sarcomeric disorganization, myofiber regeneration, and decreased force

production 140. Together, these phenotypes support that DMPK haploinsufficiency contributes to some DM1 pathologies, which could complicate RNAi therapy for several reasons: (1) wild-type and mutant DMPK alleles are identical except for the 3’ UTR

trinucleotide repeat expansion. Therefore, therapeutic RNAi would theoretically knock

down normal and mutant DMPK equally. Reducing normal DMPK could counterweigh

any beneficial effects caused by silencing the expanded mutant allele. (2) To target the

mutant allele specifically, disease-linked polymorphisms, located outside the CTG repeat

area, would have to be identified and (3) if possible, complete mutant DMPK knockdown

would yield half the normal DMPK amount, potentially resulting in haploinsufficiency-

related DM1 phenotypes. One strategy to circumvent these potential problems would

involve knocking down endogenous mutant and wild-type DMPK and simultaneously

delivering a normal DMPK cDNA engineered with base changes that prevent its

regulation by the therapeutic miRNA. A final issue regarding the feasibility of DM1-

targeted RNAi therapeutics relates to sub-cellular localization of RNAi processes and

mutant DMPK. Mutant DMPK transcripts are nucleus-sequestered, but recent

conventional thinking was that RISC-mediated gene silencing only occurred in the cytoplasm, raising doubts about whether RNAi therapy could work for DM1 141.

However, several recent studies demonstrated that nuclear RISC exists, and that RNAi

can reduce nuclear-localized transcripts, including 7SK and, importantly, DMPK 142-145.

Therefore, RNAi therapy for DM1 is feasible, but some complicating factors, discussed

above, may have to be addressed to make it a therapeutic option in humans.

31 Facioscapulohumeral muscular dystrophy. As discussed earlier, FSHD is an autosomal

dominant disorder characterized by progressive and asymmetric weakness of facial,

shoulder, and limb muscles. It was shown to arise from contracted DNA repetitive

elements though no genes are completely lost or mutated as a result of FSHD-associated

DNA deletions. Since the primary genetic mutation was identified ~17 years ago,

numerous studies have attempted to implicate specific genes in FSHD pathogenesis using

global expression analysis (e.g. RNA microarray or proteomics) or by investigating

candidate genes based on 4q35 proximity, such as DUX4, FRG1, FRG2, and ANT1

38,39,46,49,50,52,146-149. DUX4 is arguably the best FSHD candidate gene currently under investigation, however, to date, it has not been definitively accepted as the FSHD gene.

Therefore, RNAi therapy may soon be a valuable therapeutic approach for FSHD as more definitive data implicating specific genes in FSHD pathogenesis accumulates.

2.8 Summary

RNAi therapeutics is an emerging field. Several pre-clinical studies demonstrated its immense potential for treating dominant neurodegenerative diseases, chronic viral infection, and cancer. Likewise, RNAi may be a leading candidate strategy to treat dominant muscular dystrophies and other myopathies. Although currently in its infancy as a technology, its potential for disease-allele specificity may someday allow RNAi therapeutics to be a tool for personalized medicine.

32 Table 2.1: Mutations causing dominant muscular dystrophies and other myopathies

Gene Clinical Disorder OMIM MYOT LGMD1A 604103 Spheroid Body Myopathy 182920 Myofibrillar Myopathy 609200 LMNA LGMD1B 159001 Emery-Dreifuss Muscular Dystrophy 181350 CAV3 LGMD1C 601253 Distal Myopathy 607801 Rippling Muscle Disease 606072 chrom 7q LGMD1D 603511 chrom 6q23 LGMD1E 602067 chrom 7q32 LGMD1F 608423 chrom 4q21 LGMD1G 609115 PABPN1 Oculopharyngeal Muscular Dystrophy 164300 FHL1 Scapuloperoneal Amyotrophy 300695 MYH7 Laing Distal Myopathy 160500 Myosin Storage Myopathy 608358 MYH2 Hereditary Inclusion Body Myopathy 605637 DMPK Myotonic Dystrophy Type 1 160900 ZNF9 Myotonic Dystrophy Type 2 602668 TTN Tibial Muscular Dystrophy 600334 DES Myofibrillar Myopathy 601419 CRYAB Myofibrillar Myopathy 608810 ZASP Myofibrillar Myopathy 609452 FLNC Myofibrillar Myopathy 609524 BAG3 Myofibrillar Myopathy unassigned TPM3 Nemaline Myopathy 609284 ACTA1 Nemaline Myopathy 161800 Congenital Myopathy w/ Fiber Type Disproportion 255310 TPM2 Nemaline Myopathy 609285 TNNT1 Nemaline Myopathy 605355 RYR1 Central Core Disease 117000 CLCN1 Thomsen Myotonia Congenita 160800 SCN4A Paramyotonia Congenita 168300 DNM2 Myotubular (or Centronuclear) Myopathy 160150 MYF6 Myotubular (or Centronuclear) Myopathy 159991 MTMR14 Myotubular (or Centronuclear) Myopathy 160150 Congential Slow-Channel Myasthenic CHRNA1 Syndrome 601462 Congential Slow-Channel Myasthenic CHRNA2 Syndrome 601462 Congential Slow-Channel Myasthenic CHRNA3 Syndrome 601462 Congential Slow-Channel Myasthenic CHRNA4 Syndrome 601462 Continued

33 Table 2.1: Continued FRG1 FSHD? 158900 DUX4 FSHD? 158900 chrom 2p13 Welander Distal Myopathy 604454 chrom 15q Nemaline Myopathy 609273 chrom 19p13 Vacuolar Neuromyopathy 601846 Tau Sporadic Inclusion Body Myositis 147421 APP Sporadic Inclusion Body Myositis 147421 PSEN1 Sporadic Inclusion Body Myositis 147421 GSK3B Sporadic Inclusion Body Myositis 147421 ApoE Sporadic Inclusion Body Myositis 147421

Mutations in at least 29 different genes cause dominant muscular dystrophies and other myopathies. The genetic defect underlying FSHD is currently unknown, but may be caused by over-expression of one or more genes. FSHD candidates include FRG1 and

DUX4. Seven other clinically characterized myopathies have not yet been linked to specific genes. Moreover, Alzheimer’s Disease-related genes, Tau, APP, PSEN1,

GSK3B, and ApoE, are over-expressed in sporadic inclusion body myositis (IBM).

RNAi-mediated reduction of each may possibly benefit sporadic IBM disease.

34

Figure 2.1. MicroRNA biogenesis pathway. See text for details. Designed therapeutic microRNA shuttles, shRNAs, and siRNAs mimic pri-, pre-, and mature-miRNAs, respectively. Upon delivery to cells, exogenous inhibitory RNAs therefore enter the microRNA biogenesis pathway at different points, but all elicit gene silencing effects.

35

Figure 2.2. Natural microRNA sequences and structures are used to design

therapeutic inhibitory RNAs. (a) Human mir-30a. Gray and black triangles point to

Drosha and Dicer nuclease sites, respectively. Note the staggered cuts leaving 2

nucleotide 3’ overhangs. Underlined sequence indicates the mature antisense guide

strand sequence. (b) Example of designed inhibitory RNAs. Messenger RNA target sequence from E. Coli LacZ gene. Mature mir-30 sequences are replaced by

complementary LacZ-targeted inhibitory RNAs. In a miRNA shuttle, some mir-30 stem and loop sequences are maintained. The former help direct DGCR8/Drosha processing.

ShRNAs are not Drosha processed; instead the 5’ end of the hairpin is defined by the transcription start site. An siRNA is produced in vitro and designed to mimic the final mature miRNA duplex. 36

Figure 2.3. Hypothetical example of mutant allele-specific Cav-3 targeting. Wild- type (WT) and mutant T78K Cav-3 sequences are shown. The C to A Cav-3 mutation is located centrally within the miRNA sequence. This theoretical T78K-specific inhibitory

RNA also contains a secondary peripheral mutation, as discussed in the text.

37

CHAPTER 3

DUX4, A CANDIDATE GENE FOR FSHD, CAUSES P53-DEPENDENT

MYOPATHY IN VIVO

3.1 Introduction

Facioscapulohumeral muscular dystrophy (FSHD) is a complex autosomal dominant disorder characterized by progressive and asymmetric weakness of facial, shoulder, and limb muscles150. Symptoms typically arise in adulthood with most patients

showing clinical features before age 30. About 5% develop symptoms as infants or

juveniles and are generally more severely affected1,151. Clinical presentation can vary

from mild (some limited muscle weakness) to severe (wheelchair dependence).

Historically, FSHD was classified as the third most common muscular dystrophy,

affecting 1 in 20,000 individuals worldwide150. However, recent data indicate FSHD is

the most prevalent muscular dystrophy in Europe, suggesting its worldwide incidence

may be underestimated1.

Typical FSHD cases (FSHD1A; heretofore referred to as FSHD) are linked to

heterozygous chromosomal deletions that decrease the copy number of 3.3 kilobase (kb)

D4Z4 repeats on human chromosome 4q3520,149. Simplistically, normal individuals have

11-100 tandemly-repeated D4Z4 copies on both 4q35 alleles, while patients with FSHD

38 have one normal and one contracted allele containing 1-10 repeats149. In addition,

FSHD-associated D4Z4 contractions must occur on specific disease-permissive

chromosome 4q35 backgrounds152-155. Importantly, no genes are completely lost or

structurally mutated as a result of FSHD-associated deletions. Thus, although the disease

was formally classified in 1954150, and the primary genetic defect identified in 199220, the

pathogenic mechanisms underlying FSHD remain unresolved.

In leading FSHD pathogenesis models, D4Z4 contractions are proposed to cause

epigenetic changes that permit expression of genes with myopathic potential156. As a

result, aberrant over-expression of otherwise silent or near-silent genes may ultimately

cause muscular dystrophy. This model is consistent with data showing normal 4q35

D4Z4 repeats have heterochromatin characteristics, while FSHD-linked D4Z4 repeats contain marks more indicative of actively transcribed euchromatin20,21,36,37,157-159. These

transcription-permissive epigenetic changes, coupled with the observation that complete

monosomic D4Z4 deletions (i.e. zero repeats) do not cause FSHD25, support the

hypothesis that D4Z4 repeats harbor potentially myopathic open reading frames (ORFs),

which are abnormally expressed in FSHD muscles. This notion was initially considered

in 1994, when a D4Z4-localized ORF, called DUX4, was first identified21,158. However,

the locus had some characteristics of an unexpressed pseudogene and DUX4 was

therefore summarily dismissed as an FSHD candidate. For many years thereafter, the

search for FSHD-related genes was mainly focused outside the D4Z4 repeats, and

although some intriguing candidates emerged from these studies, no single gene has been

conclusively linked to FSHD development27,35,39,46,49,50,52-54,146-148,160. This slow progress

39 led to the re-emergence of DUX4 as an FSHD candidate in 2007, and several recent

findings support its potential role in FSHD pathogenesis27,52,53,55,160,161.

First, D4Z4 repeats are not pseudogenes. The DUX4 locus produces 1.7 kb and

2.0 kb full-length mRNAs with identical coding regions, and D4Z4 repeats also harbor

smaller sense and antisense transcripts, including some resembling microRNAs27,53,160.

Importantly, over-expressed DUX4 transcripts and a ~50 kDa full-length DUX4 protein

were found in biopsies and cell lines from FSHD patients27,32,52-54,160. These data are

consistent with the transcriptional de-repression model of FSHD pathogenesis. In addition, unlike pseudogenes, D4Z4 repeats and DUX4 likely have functional importance, since tandemly-arrayed D4Z4 repeats are conserved in at least 11 different placental mammalian species (non-placental animals lack D4Z4 repeats), with the greatest sequence conservation occurring within the DUX4 ORF54. Second, over- expressed DUX4 is toxic to tissue culture cells and embryonic progenitors of developing lower organisms in vivo53,55,160,161. This toxicity occurs at least partly through a pro- apoptotic mechanism, indicated by Caspase-3 activation in DUX4 transfected cells, and presence of TUNEL-positive nuclei in developmentally arrested Xenopus embryos injected with DUX4 mRNA at the two-cell stage53,55,161. These findings are consistent

with studies showing some pro-apoptotic proteins, including Caspase-3, are present in

FSHD patient muscles46,162. In addition to stimulating apoptosis, DUX4 may negatively regulate myogenesis. Human DUX4 inhibited differentiation of mouse C2C12 myoblasts in vitro, potentially by interfering with PAX3 and/or PAX7, and caused developmental arrest and reduced staining of some muscle markers when delivered to progenitor cells of zebrafish or Xenopus embryos53,55,160,161. Finally, aberrant DUX4 function was directly

40 associated with potentially important molecular changes seen in FSHD patient muscles.

Specifically, full-length human DUX4 encodes a ~50 kDa double homeodomain

transcription factor, and its only known target, Pitx1, was elevated in DUX4-over-

expressing FSHD patient muscles27,52,163. These data support the hypothesis that DUX4 catalyzes numerous downstream molecular changes, which are incompatible with maintaining normal muscle integrity.

In summary, recent studies implicated DUX4 as a leading FSHD candidate gene that is over-expressed in FSHD tissue, and generally toxic to tissue culture cells and embryonic progenitors of non-mammalian organisms, possibly through activation of downstream gene involved in apoptosis. However, the in vivo myopathic potential of

DUX4 in adult placental mammalian muscle, which most closely resembles the human

FSHD condition, has not been tested. Here, we demonstrate the in vivo myopathic potential of DUX4, using zebrafish and mice. We present evidence that DUX4 over- expression causes histological and functional features consistent with muscular dystrophy. Importantly, we show that DUX4-mediated toxicity requires DUX4 DNA binding and activation of p53-dependent apoptosis. Our comprehensive in vivo data are

consistent with the hypothesis that DUX4 over-expression contributes to FSHD.

3.2 METHODS

DUX4 epitope tagging and adeno-associated virus (AAV) production. The control

AAV.CMV.hrGFP vector contained a CMV promoter and SV40 polyadenylation signal flanked by two AAV2 inverted terminal repeats (ITRs)98. DUX4 was PCR amplified

(Pfx polymerase, Invitrogen) from pGEM/4227 using primers adding a C-terminal V5

41 epitope tag. The resultant DUX4V5 product was cloned into pCR BLUNT-II TOPO

(Invitrogen), sequence verified, then ligated into the pAAV.CMV.hrGFP proviral plasmid, replacing hrGFP. The DUX4.HOX1 DNA binding domain mutant was generated by recombinant PCR. Briefly, five key DUX4 residues responsible for making

DNA contacts, W48, F49, N51, E52 and R53 were replaced with alanines using the mutagenic primers, HOX1R 5’ –

GTGCTGCCTCAGCTGGCGTGACGCCGCATCCTGAGCCGCAATCTGGACCCTG

GGCTC – 3’ and HOX1F 5’ –

GAGCCCAGGGTCCAGATTGCGGCTCAGGCTGCGGCGTCACGCCAGCTGAGGC

AGCAC – 3’. The resultant product was then sequence verified and cloned into the AAV proviral plasmid. AAV serotype 6 (AAV6) vectors were generated by the Viral Vector

Core Facility at TRINCH. Titers were determined by quantitative PCR and reported as

DNAse Resistant Particles (DRP). AAV6.CMV.DUX4-V5, 6.1 X 1012 total DRP;

AAV6.CMV.DUX4.HOX1, 1.0 X 1013 total DRP; AAV6.CMV.hrGFP, 4.8 X 1012 total

DRP.

Western blots. For in vitro studies, HEK293 cells were transfected with expression plasmids, and proteins harvested two days later in M-PER (Thermo Scientific) with protease inhibitors (Complete, Roche). For in vivo studies, protein was harvested from

TA muscles injected with low dose AAV.DUX4 or AAV.DUX4.HOX1 vectors, 10 days post-injection. Protein yield was determined by Lowry assay (Bio-Rad). Proteins (10µg

HEK293 cell extracts; 50µg muscle extracts) were separated on 15% PAGE, transferred to PVDF membranes (Immobolin-P, Millipore), and blocked in PBS-T containing 5% non-fat dry milk. Blots from were then incubated in HRP-conjugated mouse monoclonal 42 anti-V5-primary antibody (1:10000; Invitrogen) or unconjugated β-actin control antibody

(Sigma; 1:1,000) for one-hour at room temperature. In vivo blots were also incubated

with DUX4 antibody (Y621, a gift from Y.W. Chen) overnight at room temperature

(1:1,000). Unconjugated primary antibodies were then washed and incubated with a goat

anti-mouse or goat anti-rabbit HRP-coupled secondary antibodies, respectively (Jackson

Immunochemicals; 1:100,000). Signal was detected by chemiluminescent HRP substrate

(Immobilon Western, Millipore) and exposed to autoradiography film (Amersham

Hyperfilm MP, GE Healthcare). For loading control on blots with in vivo protein extracts, PVDF membranes were incubated with Napthol Blue Black (Aldrich Chemical

Company; 1064-48-8) according to manufacturer’s instructions.

Cell death assay. Caspase 3/7 activity was measured using the Apo-ONE Homogeneous

Caspase-3/7 Assay (Promega). HEK293 cells (60,000 cells/well) were transfected as described in the text and plated simultaneously on 96-well plates. Where indicated, Bax

channel blocker, Caspase-1 inhibitor VI, or the p53-inhibitor Pifithrin α, (all from

Calbiochem), were immediately added to media at 5µM, 160 µM, 100 µM

concentrations, respectively, and media changed 4 hours later. Caspase 3/7 activity was

measured 48 hours post-transfection using a fluorescent plate reader (Spectra max M2,

Molecular Devices). Individual assays were performed in triplicate (n=6 and n=3 for

non-inhibitor and inhibitor studies, respectively) and data reported as mean caspase

activity relative to the pCINeo control. Error bars represent standard error of the mean

(s.e.m.)

Zebrafish transgenesis and histology. Tol2 fish expression plasmids and transposase

RNA were injected into one-cell stage zebrafish embryos as previously described164. 43 Clonal lines were not generated, and all phenotypes were quantified from individually

injected animals. Body morphology phenotypes were assessed by microscopy and

embryos were fixed in 4% paraformaldehyde/PBS overnight at 4°C and paraffin-

embedded. Five-micron sections were deparaffinized and rehydrated before H&E

staining. For body morphology counts, n=53 hrGFP and 49 DUX4-injected embryos.

For tissue sectioning, n= 5 representative animals per group.

Mice. All animal studies were approved by Institutional Animal Care and Use

Committee at the Research Institute at Nationwide Children’s Hospital (TRINCH).

Female C57BL/6 mice were purchased from the National Cancer Institute, Charles River and Jackson Laboratory. Trp53-/- mice165 were acquired from Jackson Laboratories

(B6.129S2-Trp53tm1Tyj/J).

AAV injections. Six to eight-week-old C57BL/6 females and eight to twelve month

Trp53-/- males and females received 8 X 108 or 3 x 1010 DRP units of AAV6 bilaterally

via direct 30 microliter intramuscular (IM) injection into the TA. In vivo transduction

was determined in AAV.CMV.hrGFP-injected using a fluorescent dissecting microscope

(MZ16FA, Leica) at x4.63 magnification.

Grip strength testing. Muscle grip strength was assessed weekly as indicated (Columbus

Instruments). Three separate trials were recorded per limb group and force measurements averaged (n=5 animals per group). Data are reported as mean hindlimb:forelimb ratios +/- s.e.m.

Electrophoretic Mobility-Shift Assay

44 Nuclear extracts were prepared from transfected HEK293 cells harvested 24 hours post- transfection. EMSA was performed as previously described52. Rabbit polyclonal V5

antibodies were used for supershifts (Chemicon).

Histological analysis. TA muscles were dissected from IM injected mice at indicated

times post-injection for histological analysis (n=6 muscles per group each time point at 8

X 108 DRP units). 10-µm cryosections were generated and H&E stained as previously

described 166. Fiber diameter and central nuclei quantifications (+/- s.e.m.) were

determined from muscles injected 2- and 4-weeks prior (n=5 muscles per group; 5 representative 20x photomicrographs per section), using AxioVision 4.7 software (Zeiss).

Fiber size-distribution histograms represent percentage of total fibers analyzed.

Immunohistochemistry. For DUX4 and Caspase 3 detection, cryosections were post- fixed in absolute methanol, blocked in PBS containing 5% normal goat serum (ZYMED laboratories), 0.1% Type A porcine gelatin (Sigma), 1% bovine serum albumin (New

England BioLabs Inc), and 0.2% Triton X-100 (Fisher Scientific), and incubated overnight at 4°C with rabbit polyclonal anti-V5 primary antibody (1:2,500; Chemicon) or rabbit polyclonal Caspase 3 primary antibody (1:1000; Abnova). After washing, sections were incubated at room temperature for 1 hour with Alexa-594 or -488 goat anti-rabbit

IgG secondary antibodies (Molecular Probes). Alternatively, V5 antibody (Chemicon) or caspase-3 antibodies (Cell Signaling) were directly labeled using the Zenon Rabbit IgG

Labeling Kit (Molecular Probes) following manufacturer’s instructions. Apoptotic nuclei were detected by TUNEL assay (In Situ Cell Death Detection Kit, Fluorescein; Roche)

45 following manufacturer’s instructions. All slides were mounted in Vectashield (Vector

Laboratories) plus DAPI.

Real time PCR array. RNA was harvested from C57BL/6 TA muscle 10 days after IM

injection of 8 X 108 DRP DUX4 or DUX4.HOX1 vectors (TRI Reagent, Molecular

Research Center Inc; followed by RNeasy mini column purification, Qiagen). N=6 muscles per group, but TA muscle RNAs from individual mice were pooled, such that n=3 for real time PCR analysis. RNA integrity was measured on a 2100 Bioanalyzer

(Agilent Technologies), quantified (NanoDrop ND-1000 Spectrophotometer), DNase

treated for 30 minutes at 37°C (DNA-free, Ambion), and reverse transcribed with random

hexamers (High Capacity cDNA Reverse Transcription Kit, Applied Biosystems).

RT2ProfilerTM Mouse Apoptosis PCR Arrays (SABiosciences) were performed and

analyzed following manufacturer’s instructions.

Statistical Analysis. All statistical analyses were performed in GraphPad Prism 5

(GraphPad Software, Inc.) using indicated statistical tests.

3.3 RESULTS

Epitope-tagged DUX4 causes apoptosis in vitro

We hypothesized that DUX4 over-expression is an underlying pathogenic insult

in FSHD. In this study, our ultimate goal was to examine the in vivo effects of DUX4

over-expression in adult mouse muscle. As an initial step, we developed DUX4

expression vectors, and validated their protein expression and cytotoxic potential in vitro.

To simplify DUX4 protein detection, we first added a C-terminal V5-epitope tag to

normal human DUX4 cDNA, and then confirmed that the V5 tag did not impact DUX4 46 expression and pro-apoptotic function in tissue culture (Fig 3.1). All references to DUX4

in experiments hereafter refer to DUX4 with a C-terminal V5 tag.

DUX4 is detrimental to developing zebrafish muscle

We next evaluated the in vivo effects of DUX4 over-expression in animal muscle.

We initially screened for DUX4 myotoxicity in zebrafish because they are rapidly produced and were previously used to investigate the pathogenesis of other human muscular dystrophies167.

We used the Tol2 transposon system and the MHCK7 promoter to generate

transgenic zebrafish with striated muscle-specific hrGFP or DUX4 expression (Fig 3.2

A)164. The MHCK7 promoter turned on 3 days post-injection, and consistent with previous reports in mice, was active in zebrafish skeletal muscle and heart (Fig 3.2 B and

Ref 168. By four days, 47% of MHCK7.DUX4-injected embryos had gross body

malformations, and 81% had strikingly abnormal muscle histology, while most

MHCK7.hrGFP-injected fish were normal (81% histology; 93% body shape; Fig 3.2

D,E). The abnormalities in MHCK7.DUX4-injected fish were consistent with those reported in other zebrafish models of muscular dystrophy167. In addition, mis-expression of DUX4 in the heart caused cardiac hypertrophy in some fish, while hrGFP did not (Fig

3.2 C). Although the expression patterns of DUX4 are unknown, cardiac defects are not normally seen in patients with FSHD, and the heart abnormalities in fish were likely artifacts of MHCK7-directed cardiac DUX4 expression. Nevertheless, these data demonstrated that striated muscles in general are susceptible to DUX4-induced myotoxicity. Importantly, in contrast to previous studies showing that ubiquitous DUX4

47 expression in developing lower eukaryotes caused embryonic arrest, our data demonstrate

that embryos expressing DUX4 specifically in muscle are viable.

DUX4 toxicity requires DNA binding in vitro

We hypothesized that DUX4-mediated apoptosis in vitro (Fig 3.1) and

myotoxicity in developing vertebrate muscle (Fig 3.2) was directly related to its ability,

as a transcription factor, to bind promoter DNA and stimulate transcription of

downstream genes. However, because in vivo over-expression of otherwise inert proteins

can sometimes be toxic to striated muscle169, we also considered the possibility that

DUX4-induced abnormalities were artifacts of non-specific over-expression unrelated to

its transcriptional activity. To rule out the latter, we generated a mutant DUX4 construct

containing alanine substitution mutations in the first DUX4 DNA binding domain, to

produce a structurally intact but functionally deficient protein (DUX4.HOX1; Fig 3.3 A).

Transfection of a CMV.DUX4.HOX1 expression plasmid into HEK293 cells produced a

DUX4.HOX1 protein migrating at the same apparent molecular weight (~50 kDa) as

normal DUX4, but with consistently higher expression levels (Fig 3.3 B). Importantly,

unlike DUX4, the DUX4.HOX1 mutant did not cause apoptosis in vitro and showed reduced binding to a DUX4-binding site in the PITX1 gene (Fig 3.3 C,D). These results

suggested DUX4 DNA binding was required to elicit pro-apoptotic effects in vitro, and

that DUX4-mediated cytotoxicity required a specific transcriptional function of DUX4.

We used the DUX4.HOX1 mutant as a control in subsequent in vivo experiments.

DUX4 causes muscle degeneration in adult mouse muscle

Since FSHD is typically an adult-onset disorder of skeletal muscle, we next

investigated the myotoxic potential of DUX4 over-expression in adult, post-mitotic

48 mouse muscle. To do this, we bilaterally injected 3 X 1010 DRP AAV6.DUX4,

AAV6.DUX4.HOX1, or AAV6.hrGFP vectors into the tibialis anterior (TA) muscles of 6 to 8-week old C57BL/6 mice. One week post-injection, we confirmed widespread TA muscle transduction by hrGFP epifluorescence and assessed histopathological changes in muscles injected with each group of vectors (Fig 3.4 A). We found that DUX4-injected muscles developed strikingly massive lesions containing degenerating myofibers and infiltrating mononuclear cells, indicating severe damage (Fig 3.4 A). In contrast, muscles

injected with hrGFP or DUX4.HOX1 controls showed no evidence of damage at identical

vector doses (Fig 3.4 A). Moreover, DUX4-injected animals were weaker than controls

1- and 2-weeks post-injection, but recovered strength by 3-weeks, which was likely due

to loss of non-integrating AAV vectors in degenerating myofibers and subsequent normal

muscle regeneration (Fig 3.4 B). Although these initial experiments demonstrated the

myopathic potential of DUX4, the massive lesions we observed at the high 3 X 1010 DRP

dose were inconsistent with more subtle and focal damage typically associated with

FSHD. We hypothesized that our high dose vector was accelerating the development of

pathology, which was reported in other vector-based models of disease170,171. We

therefore delivered lower DUX4 vector doses (8 X 108 DRP), which caused milder and less widespread muscle degeneration consistent with what is observed in FSHD patients

(Fig 3.4 C). In contrast, DUX4.HOX1 or hrGFP-transduced muscles showed no histological abnormalities even by 4 weeks post-injection, our latest time point (Fig 3.4

D). We found DUX4-induced degeneration was accompanied by significant muscle regeneration, which is a feature typical of muscular dystrophy. Specifically, 2- and 4- weeks post-injection, myofibers injected with low dose vectors had smaller mean

49 diameters, broad size variability, and dramatically increased numbers of centrally-located nuclei compared to saline-, DUX4.HOX1- or hrGFP-injected controls, which remained normal (Fig 3.4 D-F). We confirmed DUX4 and DUX4.HOX1 transgene expression by real-time PCR and immunofluorescence staining of serial muscle cryosections using rabbit polyclonal primary V5 antibodies (Figs 3.5-3.7). DUX4 and DUX4.HOX1 mRNAs and protein were expressed at similar levels (Fig 3.5), and as expected for a transcription factor163, DUX4 was primarily localized to myonuclei (Fig 3.6-3.7), but we

also found cytoplasmic DUX4 staining in degenerating myofibers (Fig 3.6-3.7). This

cytoplasmic staining was specific, since we did not see complete overlap of stained

myofibers using a second rabbit polyclonal antibody (Fig 3.7). In contrast, the

DUX4.HOX1 protein was restricted to myonuclei, and we never found myofibers

expressing cytoplasm-localized DUX4.HOX1.

DUX4 causes apoptosis through a p53-dependent mechanism

The presence of Bax and Caspase-3 proteins in affected myofibers from FSHD patients supports that apoptosis may at least partly contribute to FSHD-associated muscle wasting46,162. Because DUX4 induces apoptosis in vitro (Fig 3.1, Fig 3.3)53,55, we next

investigated the possibility that DUX4 caused cell death through an apoptotic mechanism

in adult placental mammal muscle in vivo, thereby potentially linking DUX4-mediated

cell death mechanisms to known FSHD-associated pathology. As a general screen for

apoptosis, we detected DNA fragmentation by TUNEL-staining muscle cryosections

from AAV6.DUX4- or AAV6.DUX4.HOX1-injected mice. One week post-injection, only DUX4-expressing muscles contained TUNEL-positive nuclei (Fig 3.6), which is consistent with TUNEL-positive nuclei found in DUX4-injected early Xenopus

50 embryos161. We found several TUNEL-positive nuclei that were also DUX4-positive.

Since intramuscularly delivered AAV6 preferentially transduces muscle cells but not

inflammatory mononuclear cells172,173, and we delivered our vectors to wild-type muscles lacking inflammatory infiltrates at the time of injection, we concluded that any

DUX4+/TUNEL+ nuclei were present within myofibers. Nevertheless, it is possible that

some TUNEL-positive nuclei were present within infiltrating immune cells, which

normally undergo apoptotic cell death. Moreover, TUNEL stains may also indicate

necrotic cell death. For these reasons, we more closely examined the status of apoptotic

pathways in DUX4- and DUX4.HOX1-transduced muscle, first using real-time PCR

arrays of 85 different genes involved in apoptosis (Table 3.1). We found that 36 genes

(42%) were significantly increased (>1.5-fold, p<0.05) in DUX4-injected muscles, compared to DUX4.HOX1 controls, and 12 genes (33%) were members of the p53 pathway (Table 3.2). Because DUX4-induced degeneration was associated with mononuclear cell infiltration, and inflammatory cells eventually undergo apoptosis, it is possible that these pro-apoptotic genes were changed primarily in the immune infiltrates.

To determine if apoptosis was occurring in DUX4-expressing myofibers, we stained serial cryosections from DUX4- or DUX4.HOX1-transduced muscles with antibodies to

Caspase-3, because it was the most highly up-regulated p53 pathway gene in DUX4- expressing muscle, and was previously associated with human FSHD (Table 3.2)46,162.

Muscles expressing DUX4.HOX1 were histologically normal and lacked Caspase-3

staining, while DUX4-transduced muscles contained numerous Caspase-3 positive,

degenerating myofibers (Fig 3.7). Moreover, DUX4 was expressed in essentially every

degenerating myofiber, and similar to our observations using TUNEL staining (Fig 3.6),

51 we identified differentially stained cells that were likely in different apoptotic stages.

Specifically, several myofibers expressing abundant cytoplasmic DUX4 and caspase-3

protein were nearly devoid of acidophilic eosin staining (Fig 3.7). Cells with this pattern

were likely near the terminal stages of apoptosis, since Caspase-3 activation is a late

event during the apoptotic cascade and cytoplasmic localization of the normally nuclear-

sequestered DUX4 protein suggests apoptotic nuclear breakdown. We also found

degenerating myofibers that were Caspase-3 negative but otherwise expressed DUX4 in

the nucleus and/or cytoplasm (Fig 3.7). These cells may be in early apoptotic stages

upstream of Caspase-3 activation. Finally, we identified several histologically normal

myofibers containing DUX4-positive myonuclei and no Caspase-3 expression (Fig 3.7).

The lack of complete overlap of DUX4 and Caspase-3, using two different rabbit

polyclonal antibodies, supports the specificity of our antibody stains (Fig 3.7, 3.9).

Together, our real-time PCR and immunostaining data suggested DUX4-induced cell

death occurs through a p53 pathway-dependent mechanism.

To test this hypothesis, we first determined whether p53 pathway inhibition could

prevent or blunt DUX4-induced Caspase-3/7 activation in vitro. We chose to chemically

inhibit p53, Caspase-1 and Bax because these genes are key components of signaling

cascades that ultimately lead to Caspase-3-associated apoptosis174-178, and all were activated by DUX4 over-expression in mouse muscle (Table 3.2). For this experiment, we separately pre-treated HEK293 cells with chemical inhibitors to the aforementioned genes, transfected cells with DUX4 or DUX4.HOX1 expression plasmids, and measured

Caspase-3 activity 48 hours later. Consistent with our previous findings, DUX4 alone caused significant Caspase-3/7 activation, while DUX4.HOX1 did not. Importantly, we

52 found that p53, Caspase-1, or Bax inhibition prevented or significantly reduced Caspase-

3/7 activation by DUX4 in vitro (Fig 3.8 A). To confirm these results in vivo, we injected low-dose (8 X 108 DRP) AAV6.DUX4 or AAV6.DUX4.HOX1 into TA muscles of

C57BL/6 or p53 knockout mice (Trp 53 -/-; B6.129S2-Trp53tm1Tyj/J, Jackson Labs)165,

confirmed expression by real-time PCR and immunostaining (Fig 3.5 and Fig 3.8 B), and

examined muscle histopathology 2 weeks post-injection (Fig 3.8 C). As expected, DUX4 was toxic to wild-type muscle, indicated by widespread presence of regenerating myofibers containing centrally-located nuclei (Fig 3.8 C and Fig 3.4), while muscles from p53 knockout mice appeared normal and showed no indications of the massive muscle degeneration and regeneration typified by DUX4-transduced wild-type muscles.

This is most obviously observed quantitatively by the comparatively small percentage of myofibers containing centrally located nuclei in DUX4-injected Trp53 -/- mice (4%) relative to wild-type muscles (Fig 3.7; 63%), two weeks post-injection. Moreover, the

Trp53 -/- central nuclei values do not differ significantly from control-injected wild-type muscles, and may arise from physical damage caused by the injection needle. These findings strongly support that DUX4 toxicity is p53-pathway dependent.

3.4 DISCUSSION

According to the most widely accepted pathogenesis model, FSHD is caused by genetic and epigenetic abnormalities that create a favorable environment for over- expression of genes with myopathic potential. Since 1992, several important publications described FSHD-associated genetic and epigenetic changes that are consistent with this model20,35,36,68,149,157,158,179, but the downstream transcriptional abnormalities contributing

53 to FSHD pathogenesis are unclear. This uncertainty has not arisen from a lack of

investigation; indeed, numerous groups identified potential FSHD gene candidates based

on 4q35 localization or expression changes in gene profiling experiments27,35,38,47-50,52,146-

148,158,160,180,181. However, a legitimate FSHD candidate gene should minimally satisfy

three main criteria, and to date no single gene has met each requirement. Specifically, at

a minimum, an FSHD candidate gene should: (1) show consistent over-expression in

muscles from FSHD patients; (2) have the capability to damage muscles when over-

expressed in vivo, and (3) be activated specifically in preferentially-affected FSHD

muscles (e.g. facial, shoulder-girdle, limb muscles) and/or non-muscle areas of pathology

(retina, inner ear).

DUX4 emerged as an intriguing FSHD candidate because of its position within

the D4Z4 repeats, and, importantly, several recent studies showed it was over-expressed

in affected muscles and cell lines from FSHD patients32,52,53,160. Thus, these initial

reports demonstrated that DUX4 satisfies the first criteria for an FSHD candidate gene,

although additional DUX4 expression studies in FSHD patient biopsies are required to

make this assertion more definitive. The main focus of our study was to determine

whether DUX4 satisfied the second requirement of an FSHD candidate, by possessing sufficient in vivo myopathic potential to be worthy of further study in FSHD

pathogenesis. We reasoned that the best approach to assess this was by over-expressing

DUX4 in adult mouse muscles, for three reasons. First, based on the transcriptional de- repression model of FSHD pathogenesis, it is necessary to study FSHD candidate genes by over-expressing them. Second, FSHD is typically an adult-onset muscular dystrophy

with comparatively little or no non-muscle pathology, which supports the hypothesis that

54 FSHD genes are expressed solely in areas that show pathology (i.e. primarily muscles but

perhaps also retinal vasculature and inner ear). Third, D4Z4 repeats, which are

inextricably linked to FSHD development, are only present in placental mammals,

including mice54. Thus, directing DUX4 expression in adult mouse muscle is the most

feasibly relevant model system for testing its myopathic potential in humans.

Importantly, using AAV6 vectors to deliver DUX4 to muscle, we demonstrated that

DUX4 has the potential to damage adult mouse muscle in vivo. Our findings strongly

support the hypothesis that DUX4 plays a role in FSHD pathogenesis. We also

uncovered a novel mechanism of DUX4 toxicity involving the p53 pathway (discussed in

greater detail below).

Although our study represents the first direct demonstration of the myopathic

potential of DUX4 in an adult mammalian muscle model, we are not the first to show the

harmful effects of DUX4 over-expression in general. Several recent studies

demonstrated that unregulated or ubiquitously over-expressed DUX4 was toxic to pluri-

or multi-potent progenitor cells of early Xenopus or zebrafish embryos (which lack D4Z4 repeats), or in cultured mammalian cells52,53,55,160,161. The general toxicity of DUX4

raises an important issue about its specificity, since one could argue that if DUX4 is

indeed involved in FSHD pathogenesis, its toxic effects should be restricted solely to

mammalian muscle. Thus, general toxicity could suggest DUX4 is not involved in FSHD

but is instead an artifact caused possibly by non-specific protein overload and/or

interference with the normal function of other similar homeodomain transcription factors.

Indeed, precedence exists for both possibilities: over-expression of otherwise inert GFP

protein was non-specifically toxic to striated muscle169, and at sublethal doses, DUX4

55 inhibited C2C12 cell differentiation by competing with PAX3/PAX755. Our

DUX4.HOX1 mutant ruled out the former, as high levels of this protein caused no

abnormalities in vitro or in vivo, but not the latter, since its reduced DNA binding ability would likely preclude its interference with other similar transcription factors. Regardless, the general toxicity of DUX4 does not rule out its involvement in FSHD pathogenesis.

Considering our novel findings that DUX4 activates p53-dependent cell death (Table 3.2

and Fig 3.8), its toxicity to non-muscle cells and embryos of lower organisms is not surprising, since the p53-pathway is conserved in zebrafish, Xenopus, and most non- tumor mammalian cell lines. Interestingly, the only cell line with known resistance to high levels of DUX4 protein expression is derived from a human rhabdomyosarcoma tumor typically associated with loss of p53 tumor suppressor function, which is consistent with our finding that DUX4 does not damage muscles from p53 null mice53,182.

Thus, the ability of DUX4 to activate conserved cell death pathways argues for its

contribution to FSHD if it is preferentially expressed in areas of FSHD pathology (the third criteria for an FSHD candidate gene; as yet undetermined for DUX4). Our zebrafish data demonstrate this point. Ubiquitous DUX4 over-expression was incompatible with normal zebrafish and Xenopus embryonic development in previous studies160,161, but we showed that muscle-directed DUX4 expression produced viable

zebrafish with varying degrees of dystrophic abnormalities, including asymmetrical

defects, which are a hallmark of FSHD (Fig 3.2)150. Finally, it is not unprecedented for

over-expressed disease genes, particularly ones definitively linked to a specific genetic

disorder, to cause more aggressive and widespread phenotypes in animal models than

what are typically seen in humans. For example, the most widely used mouse model for

56 Huntington Disease (HD; R6/2 model), which over-expresses a mutant human huntingtin gene, recapitulates the striatal neuron dysfunction and motor abnormalities that are hallmarks of the human disease183. Nevertheless, striatal pathology in R6/2 mice is

significantly more aggressive than in humans, and they also display widespread

phenotypes that are not present in typical HD patients, including reduced fertility, very

early death, epilepsy, diabetes, cardiac dysfunction, and neuromuscular junction defects183. More recently developed models that more faithfully genocopy human HD do

not display such widespread abnormalities183. This case study analogy illustrates that

even a definitive human disease gene, like HD, can cause unexpected widespread toxicity

in animal models. Our data strongly demonstrate the in vivo myopathic potential of

DUX4; considering our discussion above, our study further supports the hypothesis that

DUX4 over-expression is an underlying pathogenic event in FSHD. In future studies, it

will be important to address the third criteria of an FSHD candidate gene by testing the

myopathic potential of DUX4 using natural human D4Z4-derived promoter and polyA

elements.

We also reported the novel finding that the pro-apoptotic effects of DUX4 were

p53-dependent (Table 3.2, Fig 3.8). Our data therefore support a mechanism for DUX4

pro-myopathic activity, which could also explain its general toxicity to most cells in

which it is over-expressed. Although previous discoveries that p53 pathway components

(Bax and Caspase-3) are activated in FSHD muscles is consistent with our DUX4/p53

findings, it will be important in future studies to better define p53 pathway involvement

in FSHD. In addition to searching for p53 pathway activation in FSHD muscles and cell

lines, additional mechanistic data are needed to determine whether DUX4 directly

57 activates the p53 promoter or does so indirectly through activation of intermediary gene

products. One potential mechanism for the latter may involve DUX4 activation of the

PitX1 gene, which activates p53 and is increased with DUX4 in FSHD patient

biopsies52,184.

Although DUX4 is clearly a transcription factor, its normal biological roles are

unknown. Here, we reported that DUX4 activates the p53 pathway and Caspase-3, which play important roles in skeletal muscle differentiation and regeneration185,186. It is

therefore possible that DUX4 normally functions to regulate skeletal muscle

development, but in FSHD, its over- or mis-expression negatively impacts muscle development and regeneration through chronic p53-pathway activation and/or interference with other homeodomain transcription factors (like PAX3/PAX7). If DUX4 is expressed in muscle progenitors, as suggested55, chronic DUX4-induced damage could reduce the pool of proliferating satellite cells, causing regeneration defects that could, over time, manifest as the progressive weakness that typifies FSHD. Thus, understanding where and when DUX4 is normally expressed in vivo will help further define its potential mechanistic role in FSHD pathogenesis.

Finally, our data do not rule out the involvement of other FSHD-associated potentially myopathic genes39,148, independent of DUX4. It is possible that DUX4 over- expression is one of multiple pathogenic insults that conspire to produce FSHD pathologies 39. Nevertheless, our data show that DUX4 is highly toxic to mammalian

muscle, at least partly through activation of p53-dependent cell death. Since it is a

transcription factor, even small perturbations in its expression could dramatically alter

p53 signaling or expression of other genes required to maintain normal muscle integrity.

58 Thus, additional characterization of DUX4-controlled pathways, including p53, may help define mechanisms contributing to muscular dystrophy and ultimately provide targets for therapeutic intervention of FSHD.

59 Table 3.1. Total apoptotic gene expression changes in DUX4-injected muscles

Gene Symbol Fold Change P value Significantly changed genes Hells 14.86 0.0068 Tnf 13.15 0.0096 Il10 12.97 0.033 Bcl10 9.83 0.018 Naip2 8.13 0.0045 Casp3 7.87 0.013 Bcl2l1 7.13 0.0042 Birc5 6.98 0.0068 Cd40 6.22 0.024 Pycard 5.92 0.015 Akt1 4.88 0.0064 Bax 4.86 0.0082 Casp1 4.85 0.034 Apaf1 4.44 0.016 Bok 4.32 0.022 Casp6 4.09 0.042 Ltbr 3.98 0.0069 Casp4 3.93 0.045 Trp63 3.93 0.0035 Bid 3.78 0.0022 Casp9 3.76 0.0097 Nfkb1 3.66 0.019 Bak1 3.40 0.030 Fas 3.23 0.025 Hsp90ab1 3.12 0.0093 Trp53 3.06 0.029 Api5 2.79 0.023 Bad 2.68 0.036 Tnfrsf1a 2.66 0.020 Casp7 2.57 0.0065 Traf3 2.45 0.039 Sphk2 2.13 0.013 Prdx2 2.11 0.039 Cflar 2.06 0.032 Bag3 1.91 0.022 Mcl1 1.73 0.013 Unchanged or insignificantly changed genes Atf5 1.12 0.43 Bag1 1.35 0.16 Bcl2 1.63 0.19

Continued

60 Table 3.1. continued Bcl2l10 -2.82 0.087 Bcl2l2 2.21 0.068 Birc2 1.21 0.15 Birc3 1.71 0.18 Bnip2 1.91 0.058 Bnip3 -1.15 0.44 Bnip3l -1.02 0.91 Card10 1.31 0.10 Card6 1.19 0.53 Casp12 1.31 0.50 Casp14 1.91 0.095 Casp2 -1.06 0.83 Casp8 1.05 0.90 Cd40lg -2.38 0.22 Cd70 -1.74 0.24 Cidea 1.13 0.85 Cideb 1.62 0.15 Cradd -1.16 0.55 Dad1 1.54 0.48 Dapk1 1.60 0.23 Dffa 1.26 0.28 Dffb 1.58 0.23 Fadd 1.10 0.90 Fasl -1.74 0.24 Lhx4 -1.32 0.58 Naip1 -1.05 0.93 Nme5 1.01 0.97 Nod1 -1.40 0.32 Nol3 1.07 0.63 Pak7 -1.59 0.34 Pim2 -1.47 0.12 Polb 1.14 0.69 Ripk1 1.93 0.12 Rnf7 1.71 0.089 Tnfrsf10b 1.48 0.34 Tnfrsf11b 1.54 0.18 Tnfsf10 -1.25 0.33 Tnfsf12 1.95 0.064 Traf1 1.64 0.21 Traf2 1.37 0.20 Trp53bp2 1.11 0.73 Trp53inp1 1.18 0.48 Trp73 1.92 0.093 Tsc22d3 2.26 0.19 Xiap 1.50 0.11 Zc3hc1 2.03 0.064

61 Table 3.2. Significantly Changed p53 Pathway Genes in DUX4-Injected Muscles

Gene Symbol Fold Change p Value Casp3 7.87 0.0130 Birc5 6.98 0.0068 Bax 4.86 0.0082 Casp1 4.85 0.0340 Apaf1 4.44 0.0160 Trp63 3.93 0.0035 Bid 3.78 0.0022 Casp9 3.76 0.0097 Bak1 3.4 0.0300 Trp53 3.06 0.0290 Bad 2.68 0.0360 Casp7 2.57 0.0065

62

Figure 3.1. V5 epitope tagged DUX4 caused apoptosis in vitro. (a) To simplify DUX4 protein detection, we first added a C-terminal V5-epitope tag to normal human DUX4 cDNA by PCR and cloned the resultant product into a mammalian expression vector used in all cell and mouse experiments. ITR, AAV inverted terminal repeat; CMV, cytomegalovirus promoter; pA, SV40 poly A signal. (b) Following transfection of expression plasmids into HEK293 cells, we confirmed DUX4 protein expression by western blot using V5-epitope antibodies. (c) DUX4 was previously shown to induce apoptosis in vitro. We validated that the V5 tag did not impact DUX4 pro-apoptotic function by measuring Caspase 3/7 activity in HEK293 cells transfected with plasmid expressing DUX4.V5, untagged DUX4, FRG1, and mock or empty vector controls. *, indicates significantly increased Caspase 3/7 activity compared to mock lipofectamine or empty CMV vector (pCINeo) controls, or CMV.FRG1, p<0.0007 (ANOVA). 63 Figure 3.2. DUX4 over-expression is detrimental to developing zebrafish muscle.

(a) Tol2 zebrafish expression constructs contained striated muscle-specific MHCK7

promoter-driven DUX4 or hrGFP. ITR, inverted terminal repeat from Tol2 transposon.

pA, SV40 polyA signal. (b) hrGFP epifluorescence showed MHCK7 activity in zebrafish

muscle, which turned on three days post-injection. This lag in MHCK7 promoter

expression allowed culling of abnormal embryos arising from non-specific plasmid

toxicity within the first 2 days post-injection. (c) MHCK7.DUX4 caused body

malformation defects including short anterior-posterior (AP) axes, curved bodies,

asymmetrically undeveloped pectoral fins (arrow indicates fin), or combinations of these

morphologies. Some fish also showed cardiac hypertrophy (asterisk indicates heart) due

to MHCK7-mediated DUX4 expression in the myocardium. (d) Hematoxylin and eosin

(H&E) staining of zebrafish body muscle shows MHCK7.DUX4 zebrafish had undefined somite boundaries, absent sarcomeric banding, and myofiber disorganization/degeneration. In contrast, MHCK7.hrGFP had no significant impact on gross body formation, or somite/myofiber organization compared to normal zebrafish embryos. The only abnormal phenotypes seen in hrGFP fish were short AP axes, while undeveloped pectoral fins and abnormal body shapes were never present. Scale bars, 50

µm. (e) Quantification of abnormal muscle phenotypes in zebrafish pictured in c-d. All

DUX4-injected fish with abnormal body morphology also showed histological defects.

64

Figure 3.2

65 Figure 3.3. DUX4 toxicity requires DNA binding. (a) Structure of DUX4 AAV expression construct. White boxes indicate homeodomains (labeled 1 and 2). ITR, AAV inverted terminal repeats. CMV, cytomegalovirus promoter. pA, SV40 polyA signal.

Alignment with a consensus homeodomain (Cons. Hox) identified 5 important residues required for DNA binding. *, indicates residues mutated to alanines in DUX4.HOX1

DNA binding mutant. (b) Western blot using extracts from transfected HEK293 cells showed DUX4.HOX1 protein was expressed at expected molecular weight (~50 kDa) and consistently produced at higher levels than normal DUX4, in vitro. Lipo, indicates

HEK293 cells transfected with Lipofectamine-2000™ (Invitrogen) but no DNA. (c)

Unlike DUX4, the DUX4.HOX1 mutant did not cause apoptosis in vitro, as indicated by lack of caspase-3/7 activation following transfection into HEK293 cells. ***, indicates significant differences from lipofectamine controls, p<0.0001 (ANOVA; n=3 independent experiments performed in triplicate). RFU, relative fluorescent units from

Caspase-3/7assay. (d) Electrophoretic mobility shift assay (EMSA). Lanes 1 and 2

(DUX4 and DUX4 SS, respectively) show a shifted and super-shifted oligonucleotide

(oligo) corresponding to the DUX4 binding site in the PITX1 promoter52. Lane 3,

DUX4.HOX1 has lower affinity for PITX1 promoter oligo. Mutation of the PITX1

binding site (mutPITX1) further reduces binding by DUX4, as previously reported, while

no binding occurs between DUX4.HOX1 and the mutPITX1 site. Arrow, indicates free

PITX1 promoter probe; arrowhead, indicates DUX4-bound PITX1 promoter sequence;

asterisk, indicates DUX4-PITX1 promoter complex supershifted with V5 antibody. SS,

supershift.

66

Figure 3.3 67 Figure 3.4. DUX4 is toxic to adult mouse muscle in vivo. (a) hrGFP epifluorescence shows AAV6 transduction of adult TA mouse 1 week post-injection. H&E staining shows DUX4 caused massive myofiber degeneration and mononuclear cell infiltration that was not present in DUX4.HOX1 or hrGFP controls at high vector dose (3 X 1010

DRP). Scale bars, 500 µm. (b) DUX4, but not DUX4.HOX1 or controls, significantly

reduced TA muscle grip strength 1 and 2 weeks post-injection. N=5 mice per group.

(***, p<0.001; **, p<0.01; 2-way ANOVA with Bonferroni post-hoc test). H:F ratio

indicates hindlimb (transduced) to forelimb (untransduced) grip strength ratios. (c) At

lower doses (8 X 108 DRP), DUX4 caused myofiber degeneration (by one week, shown

here) that recapitulated the focal dystrophic lesions seen in FSHD patients. Arrows point

to degenerating myofibers, indicated by loss of acidophilic staining in H&E stains. Scale

bars: left panel, 500 µm; right panel, 50 µm. (d) H&E staining revealed abundant

centrally-located nuclei and myofiber size variability only in DUX4 injected muscles. PI,

indicates post-injection. Scale bar, 50 µm. (e) Distribution of fiber diameter as a

percentage of total fibers counted during sampling. DUX4 transduced muscles had more

small-bore fibers compared to all controls, which is characteristic of regenerating

dystrophic muscle. (f) DUX4-injected muscles had significantly higher percentages of

centrally-located nuclei (%C.N) at both time points, which is another feature of

dystrophic muscle, p<0.001 (Chi-square). All injections for panels c, d, and e delivered 8

x 108 DRP of AAV6 vectors.

68

Figure 3.4 69 Figure 3.5. In vivo DUX4 mRNA and protein expression. (a) Real-time PCR shows

DUX4 and DUX4.HOX1 were expressed at equivalent levels in vivo. N=4 muscles/group. Differences are not statistically significant (ANOVA). (b)

Representative western blots showing DUX4 and DUX4.HOX1 proteins were expressed

at equivalent levels in vivo. Each lane contains muscle protein extracts harvested 10 days

post-injection from low dose AAV.DUX4.HOX1 or AAV.DUX4-injected animals.

Numbers 516, 517, 518, 153, 520, 521 indicate individual mouse identification numbers.

HEK293 DUX4 lane contains protein extract harvested from HEK293 cells transfected with a CMV.DUX4V5 expression plasmid, for use as a positive control. Blots were incubated with anti-V5 or anti-DUX4 primary antibodies. Multiple DUX4 and

DUX4.HOX1 products were evident, but predominant bands migrated at 50 kDa and ~38 kDa, both of which were previously reported DUX4 protein isoforms2,4. The smaller

sections in this panel are cropped around the 50 kDa marker and digitally enhanced to

better show products corresponding to full-length DUX4 protein size. The DUX4-

specific antibody consistently produced weaker bands compared to the V5 antibody. The

bottom panel shows the membrane stained with Napthol Blue Black, to demonstrate

equal protein loading.

70

Figure 3.5

71 Figure 3.6. DUX4-transduced myofibers are TUNEL-positive. (a) Low power photomicrographs demonstrating co-localization of TUNEL-positive nuclei in DUX4-, but not DUX4.HOX1- transduced myofibers. DUX4 was localized in predominantly in nuclei, but also in the cytoplasm of degenerating myofibers. Arrows indicate nuclear- localized DUX4 present within normal myofibers. Caret indicates DUX4+/TUNEL+ myonuclei present within overtly normal myofibers. Asterisk specifies TUNEL-negative degenerating myofibers containing cytoplasmic DUX4. Arrowheads point to

DUX4+/TUNEL+ degenerating myofibers. (b) High power image of DUX4+/TUNEL+ degenerating myofiber in (a). Scale bar, 50 µm.

72 Figure 3.7. DUX4-transduced myofibers are Caspase-3 positive. Top panels show

DUX4+/caspase-3+ degenerating myofibers indicated by arrow, and shown in higher power in middle panels. Rabbit V5 antibody stain shows DUX4 was present in the nucleus but also had cytoplasmic localization in degenerating myofibers. In contrast,

DUX4.HOX1 protein was exclusively nuclear. Some degenerating myofibers were

Caspase-3 negative but expressed DUX4 in the nucleus (caret) or cytoplasm (arrowhead).

In contrast, several normal myofibers were DUX4+/Caspase-3-negative (pound sign).

Bottom panels, Caspase-3 staining was absent in histologically normal muscle expressing

DUX4.HOX1. The rabbit polyclonal Caspase-3 primary antibody used here (Abnova;

PAB0242) detects total Caspase-3. Antibodies specifically recognizing cleaved Caspase-

3 showed similar staining patterns, as demonstrated in Fig 3.9. DAPI, 4',6-diamidino-2- phenylindole, stains nuclear DNA. Scale bars, 50 µm.

73 Figure 3.8. DUX4 causes apoptosis through a p53-dependent mechanism. (a) DUX4-

induced apoptosis is significantly reduced by Bax, p53, or caspase-1 inhibition, in vitro.

**, p<0.01; ***, p<0.001 (ANOVA). RFU, relative fluorescent units from Caspase-3/7

assay. (b) V5 immunofluorescence and DAPI staining showed DUX4 expression in

Trp53 -/- mouse myonuclei 2 weeks after injection. Scale bar, 50 mm. (c) Trp53 -/-

muscles are resistant to DUX4-induced degeneration, indicated by normal muscle histology in DUX4-transduced muscles, two weeks post-injection. In contrast, low-dose

AAV6.DUX4 (8 X 108 DRP) caused massive myofiber degeneration and subsequent

regeneration two weeks post-injection. Scale bar, 500 mm.

74

Figure 3.8

75 Fig 3.9. DUX4-transduced myofibers express cleaved Caspase-3. (a-d) show low power H&E, caspase-3 (green), DUX4 (red), and DAPI (blue) stained serial sections from DUX4-transduced mouse muscle 10 days post-injection. Caret, asterisk, and pound sign serves as landmarks for orientation of high power images in panels e-l. (e-f) shows nuclear DUX4 in histologically normal myofiber (caret). (g-h) shows Caspase-3 positive myofibers containing nuclear-localized DUX4 (pound sign). (i-l) Adjacent to landmarked myofibers (indicated by asterisk and pound sign) are examples of Caspase-3 positive myofibers containing nuclear- and cytoplasm-localized DUX4. Some mononuclear cells may also be caspase-3 positive. Antibody used in this figure specifically detected cleaved caspase-3 (9662, Cell Signaling Technology). Scale bars,

50 µm.

76

CHAPTER 4

RNA INTERFERENCE IMPROVES MYOPATHIC PHENOTYPES IN MICE

OVER-EXPRESSING FSHD REGION GENE 1 (FRG1)

4.1 Introduction

The concept of muscle gene therapy arose soon after dystrophin mutations were

identified as the underlying cause of Duchenne muscular dystrophy (DMD) in 1987 187.

Because DMD was a recessive disorder caused by the lack of normal dystrophin in

muscle, several research groups began developing dystrophin gene replacement strategies

as potential treatments for DMD 188. For many years, this was the sole focus of the

nascent muscle gene therapy field, but as mutations in other myopathy-related genes were subsequently identified, the field expanded beyond DMD to include other muscle disorders 189. These disease gene identification studies, combined with important

advancements in adeno-associated viral (AAV) vector development and delivery over the

last two decades, contributed to several successful pre-clinical gene therapy trials in

animal models of various myopathies 166,188,190-198. Importantly, one recently translated

study showed the first promising Phase I clinical trial data for gene therapy of Limb

Girdle Muscular Dystrophy (LGMD2D) in humans 199. Thus, steady progress in a

77 relatively short period of time supports that gene therapy may someday be an effective

method for treating inherited disorders of muscle.

Nevertheless, most of the current progress in the field has been primarily directed

toward developing therapies for recessive disorders, while approaches to treat dominant

myopathies were largely unexplored by comparison 166,188,190-196,198. This disparity in

research focus is significant, as two of the three most common muscular dystrophies are

dominant (facioscapulohumeral muscular dystrophy, FSHD; myotonic dystrophy, DM),

and more than half of all currently known myopathy-related disease genes are linked to

dominant disorders 200. One reason the muscle gene therapy field principally focused on

recessive myopathies relates to the technical feasibility of the strategies necessary to treat

each class of disorders. Specifically, recessive disorders require gene replacement, while

dominant diseases would potentially benefit from disease gene silencing 200. Historically,

feasible molecular tools existed to accomplish the former, but not the latter. This

disparity could change however, with the recent emergence of RNA interference (RNAi)

as a promising therapeutic approach to silence dominant disease genes 201. The initial

work in this area mostly focused on treating neurodegenerative disease 201, but we

hypothesized that RNAi could also be an effective mechanism to silence genes associated

with dominant myopathies, which has not been previously illustrated 200. The goal of this

study was to demonstrate proof-of-principle that RNAi-based gene therapy could correct muscle abnormalities in a mouse model of dominant myopathy. To do this, we used the

FRG1-high transgenic mouse line, which develops myopathy caused by muscle-specific

over-expression of the human Facioscapulohumeral muscular dystrophy (FSHD) Region

Gene 1 (FRG1)39. As the gene name suggests, FRG1-high mice were initially developed as

78 a putative model of autosomal dominant FSHD, but the pathogenic mechanisms

underlying this disorder are not altogether resolved; indeed, recent data support a model

in which DUX4 over-expression is a primary pathogenic insult underlying FSHD

32,35,39,52,53,55,56. Thus, both FRG1 and DUX4 may be candidate targets for RNAi therapy,

but there are no published animal models stably expressing the latter. We therefore

focused on FRG1 in this study, and tested the potential of AAV-delivered, FRG1-targeted microRNAs to correct myopathy in FRG1-high mice. Our results demonstrate the

therapeutic promise of RNAi therapy for FSHD candidate genes specifically, and

dominant myopathies in general.

4.2 Materials and Methods

Cloning of FRG1-targeted microRNAs. Mouse U6 promoter-driven artificial

microRNAs targeting human FRG1 (called miFRG1s) were cloned using common

molecular techniques as previously described 105. All microRNAs were based on human

mir-30 sequences and structure, but the mature mir-30 portions were replaced by

sequences derived from the FRG1 coding region. Ten different miFRG1s were

generated; nomenclature indicates the first position of the miRNA binding site relative to

+1 of the FRG1 coding region. Control U6-driven microRNAs targeting eGFP (miGFP)

and firefly luciferase (miLuc) were previously described 103,105.

Luciferase assay. The luciferase reporter plasmid (Fig 1a) was modified from Psicheck2

(Promega). Human FRG1 cDNA was cloned downstream of the Renilla luciferase stop

codon, thereby functioning as a 3’ UTR. A separate TK promoter driven firefly

luciferase cassette served as a transfection control. HEK293 cells were co-transfected in

79 triplicate wells (Lipofectamine-2000, Invitrogen) with the luciferase.FRG1 reporter and

individual U6.microRNA expression plasmids in a 1:5 molar ratio. FRG1 gene silencing

was determined by measuring Renilla and firefly luciferase activity (Dual Luciferase

Reporter Assay System, Promega) 48 hours post-transfection, following manufacturer’s

instructions. Triplicate data were averaged, and individual experiments performed three

times; results were reported as the mean ratio of renilla to firefly activity ± standard error

of the mean (s.e.m.)

Real-time PCR and western blot. For in vitro work, U6.miFRG1 or control

microRNA plasmids were co-transfected with a CMV.FRG1 expression vector into

HEK293 cells (5:1 molar ratio). Forty-eight hours later, RNA or protein was extracted

(Trizol from Fisher and M-PER from Pierce, respectively). For in vivo work, RNA or protein was extracted from muscles injected 11-14 weeks prior, using previously described methods (23). RNA was quantified by Nanodrop, DNase-treated (DNA-Free,

Ambion), and reverse transcribed using random hexamers (Applied Biosystems cDNA

Archive Kit). Subsequent cDNA samples were then used as template for Taqman Assay using pre-designed FRG1 and human β-actin or mouse GAPDH control primer/probe sets

(Applied Biosystems). Two independent experiments were performed, with each sample

run in triplicate. All in vitro data were normalized to miLuc-expressing samples. For

westerns, protein was quantified by Lowry assay (BioRad), 50 µg samples were

separated on 15% SDS-PAGE, transferred to PVDF membrane, and incubated with the

following antibodies: commercial primary mouse monoclonal antibodies to FRG1

(1:8,000 Abnova); custom polyclonal FRG1 antibodies kindly provided by Dr. Peter

Jones (DMA-AP-1, 1:500)202; mouse monoclonal β-actin antibodies (1:60,000, Sigma); 80 or rabbit polyclonal α-tubulin antibodies (1:5,000; Abcam) overnight at 4°C. Following

washes, blots were then probed with HRP-coupled goat anti-mouse or goat anti-rabbit secondary antibodies (1:100,000; Jackson ImmunoResearch) for 1 hr at room temperature and then developed using Immobilin Western HRP substrate (Millipore).

AAV vector delivery to mouse muscle. U6.miGFP and U6.miFRG1.948 were cloned into our AAV.CMV.hrGFP proviral plasmid upstream of CMV.hrGFP. AAV6 particles were generated and titrated as previously described by the Viral Vector Core Facility at

The Research Institute at Nationwide Children’s Hospital 56. FRG1-high colonies were

maintained by breeding hemizygous FRG1-high mice to C57BL/6 animals. Male FRG1-

high and negative littermates were identified by PCR genotyping of genomic DNA from

newborn mice (P1 or P2) using primers detecting the HSA.FRG1 transgene (5’-

CCAGGGTAAAAAGACCATTGTCG-3’ and 5’-TCGTGCTCAAGGGAACCAAG-3’)

and the mouse Y chromosome (SRY gene; 5’-GTGTCACAGAGGAGTGGCATTTTAC-

3’ and 5’- TTGCTGCTGGTGGTGGTTATGG-3’). Following genotyping, male P1 or

P2 mice were injected in the lower limbs with 5 X 1010 DNAse Resistant Particles (DRP)

per leg with indicated vectors. All mouse procedures were performed following guidelines approved by the Institutional Animal Care and Use Committee (IACUC) at the

Research Institute at Nationwide Children’s Hospital.

Imaging and histology. In vivo AAV transduction was determined by hrGFP

epifluorescence using a fluorescent dissecting microscope (MZ16FA, Leica) at x4.63

magnification. Dissected muscles were placed in O.C.T. Compound (Tissue-Tek), frozen

in liquid nitrogen-cooled isopentane, cut onto slides as 10 µm cryosections, and stained

with hematoxylin and eosin (H&E; following standard protocols), Oil Red O/Harris 81 hematoxylin, or DMA-AP-1 FRG1 polyclonal antibodies. Oil Red O stains were

performed using a filtered 60% stock solution in dH2O (stock, 2.5g Oil red O powder in

500ml isopropanol). Cryosections were post fixed in 10% formalin for 10 min, washed in

tap water, and stained in Oil red O working solution for 10 min. Slides were then washed

in tap water, counter stained in Harris hemotoxylin for 1 min, rinsed, blued in ammonia

water, and washed in tap water. H&E and Oil Red O sections were covered with crystal-

mount (Electron Microscopy Sciences), and mounted with Permount (Fisher Scientific).

For FRG1 immunohistochemistry, cryosections were post-fixed in 4% paraformaldehyde, washed, blocked in 5% milk/PBS-tween (PBST), incubated overnight at 4°C with DMA-

AP-1 FRG1 primary antibody (1:200 in 1% BSA, 20% goat serum, and PBS), and then with AlexaFluor-594 conjugated goat anti-rabbit secondary antibodies (1:500; 1 hr at RT;

Molecular Probes). Slides were covered in Vectashield plus DAPI. All images were taken from mouse tissue harvested from 11-14 week old male mice, except in (Fig 4.4, 3 week old mice). Muscle cross-sectional fiber diameters and percentage of myofibers with centrally-located nuclei were determined as previously described from five different animals per group (5 fields per leg)56.

Grip strength. Hindlimb grip strength was measured weekly between 6-10 weeks of age

as previously described (n=8 male animals per group)56. Data represent means ± s.e.m.

4.3 Results

Myopathy in FRG1-high mice results from muscle-specific over-expression of the human FRG1 coding region 39. We therefore hypothesized that FRG1 knockdown using

RNA interference (RNAi) would improve myopathic phenotypes. To do this, we 82 designed 10 different U6 promoter-driven artificial microRNAs targeting sequences in the human FRG1 coding region (Fig. 4.1). We then identified our lead FRG1-targeted microRNA (miFRG1) using in vitro screening assays. First, we co-transfected

U6.miFRG1 constructs with a plasmid expressing a Renilla luciferase-FRG1 fusion transcript and a separate firefly luciferase transfection control (Fig. 4.2 a). We then measured miFRG1-mediated gene silencing indirectly by determining the ratio of Renilla to firefly luciferase activity from transfected cell lysates, 2 days later (Fig. 4.2 b). To confirm these gene silencing data against a normal FRG1 open reading frame, we co- transfected individual U6.miFRG1 plasmids with a CMV.FRG1 expression vector into

HEK293 cells and measured FRG1 transcript and protein levels by real-time PCR and western blot, respectively (Fig 4.2 c-d). We identified two different miFRG1 sequences that consistently silenced FRG1 using all three assays, and we chose one of these

(miFRG1.948; heretofore referred to as miFRG1) for in vivo studies because it catalyzed slightly better silencing at the protein level (Fig 4.2 d).

We next cloned U6.miFRG1 or an U6.miGFP control microRNA into our

AAV.CMV.hrGFP proviral vector 98. The miGFP control microRNA, which targets

sequences in the eGFP gene, does not direct knockdown of FRG1 or our hrGFP reporter.

We then made AAV6 viral vectors expressing hrGFP alone (AAV.hrGFP), or hrGFP

with miFRG1 or miGFP (AAV.miFRG1 or AAV.miGFP, respectively; Fig 4.3 a), and injected 5 x 1010 DNAse resistant particles (DRP) of each vector into the lower limbs of

newborn FRG1-high or wild-type male littermates. For all wild-type mouse injections, and

FRG1-high mice injected with AAV.miFRG1 vectors, this delivery approach produced

robust and widespread hrGFP expression in most major muscles of the lower limbs,

83 including the adductors (add), gastrocnemius (gas), tibialis anterior (TA), and gluteus

maximus (glut), up to 14 weeks post-injection (Fig. 4.3 b-c). The quadriceps muscle was inconsistently hrGFP positive, and showed the least amount of transduction of all major lower limb muscles (Fig 4.3 c). In contrast, we did not observe abundant hrGFP expression in AAV.hrGFP or AAV.miGFP-injected FRG1-high mice 14 weeks after injection, although it was present at 3 weeks (Fig. 4.4).

We next examined muscle size in AAV.miFRG1- and control-treated mice, since muscle size deficits are the most obvious gross abnormality in FRG1-high animals. We

found that AAV.miFRG1-treated lower limb muscles were visually larger than

AAV.miGFP-injected controls, and isolated add, gas, and glut muscles from the former

weighed significantly more than those from animals that received AAV.miGFP or

AAV.hrGFP vectors (~2-fold average increase; Fig. 2d-e). Moreover, AAV.miFRG1-

injected FRG1-high add, gas, and glut muscles were indistinguishable in size from wild-

type controls (Fig. 4.3 d-e). We observed similar trends in TA muscle sizes of all treated

and control groups, but none of these changes were statistically significant. In contrast,

AAV.miFRG1 treatment did not restore FRG1-high quad muscles to wild-type sizes. This

insignificant ~1.3-fold mean quadriceps size correction was likely due to low average

transduction, since quad weights trended higher in individual FRG1-high mice with the

most AAV.miFRG1-derived hrGFP expression (Fig. 4.3 d-e and Fig 4.5).

We next correlated these gross muscle improvements with FRG1 mRNA and protein knockdown using real-time PCR, western blot, and immunofluorescence staining.

We measured a statistically significant ~55% reduction of over-expressed human FRG1 mRNA in FRG1-high muscles treated with AAV.miFRG1 compared to AAV.miGFP

84 controls (Fig 4.6 a). AAV.miFRG1 also reduced endogenous mouse FRG1 mRNA by

~50%, despite having one mismatch with this transcript (Fig 4.1). Expectedly, total

FRG1 mRNA knockdown corresponded to a marked decrease in FRG1 protein levels by

western blot (Fig. 4.6 b), but since RNAi rarely produces complete knockdown of

abundant targets like FRG1 in FRG1-high mice, we were surprised to note that human

FRG1 was nearly or completely undetectable in moderately- to highly-transduced myofibers (Fig. 4.6 c). This knockdown in AAV.miFRG1-transduced FRG1-high muscles

was associated with wild-type histology. Specifically, all wild-type groups and transduced muscles from AAV.miFRG1-treated FRG1-high mice lacked the fibrosis, fat

deposition, and myofiber degeneration and regeneration (indicated by increases in

myofibers with smaller diameters and/or centrally-located nuclei) seen in untreated or

control-treated FRG1-high animals (Fig. 4.7).

Finally, we determined whether FRG1 knockdown in FRG1-high mice improved

overall hind limb muscle function. To do this, we measured grip strength weekly in

AAV.miGFP-, AAV.miFRG1-, or AAV.hrGFP-injected FRG1-high and wild-type male mice for 5 weeks, and compared these values to uninjected wild-type littermates. We found that AAV.miFRG1-treated FRG1-high animals were significantly stronger (~1.7- fold average increase) than age-matched AAV.miGFP- or AAV.hrGFP-injected controls

(Fig. 4.8 and Fig. 4.9). In contrast, hind limb grip strength from AAV.miFRG1-treated

FRG1-high and WT mice, and all injected wild-type animals, was not significantly different from uninjected wild-type controls (Fig. 4.8 and Fig. 4.9). We therefore

concluded that RNAi-mediated knockdown of FRG1 improved myopathy in individual

myofibers, isolated muscles, and whole limbs from FRG1-high mice.

85

4.4 Discussion

Gene therapy refers to a therapeutic approach for disease that uses nucleic acids instead of drugs (http://www.asgct.org/about_gene_therapy/defined.php). For many years, this definition was almost exclusively synonymous with gene replacement, the strategy typically used to treat recessive diseases. As RNAi and microRNA-based expression systems emerged in recent years, the gene therapy field evolved to include gene silencing as another possible therapeutic approach. Thus, the number of diseases potentially treatable with nucleic acid therapies expanded. We viewed this expansion as an opportunity to begin developing gene therapies for dominant myopathies, which historically has been an underrepresented area of research.

The ‘dominant myopathies’ classification refers to a diverse group of clinically distinct, currently incurable, and potentially devastating muscle disorders caused by mutations in at least 29 different genes 200. As a group, dominant myopathies are relatively abundant, possibly affecting as many as 1 in 2,400 to 1 in 3,200 individuals

1,200. We hypothesized that a common RNAi-based therapeutic strategy, with modifications depending on etiology of each disorder, could potentially benefit a large population of patients affected by dominantly inherited muscle disease. We therefore set out to demonstrate proof-of-principle for this strategy in vivo. Accomplishing this required a disease animal model that developed obvious myopathic phenotypes arising from expression of a gene linked to a dominant human myopathy. We used the FRG1-high mouse model in this study, which was initially developed to test the hypothesis that

FRG1 over-expression was a primary pathogenic insult underlying Facioscapulohumeral

86 muscular dystrophy (FSHD)35,39. Although the progressive myopathy produced in these

mice strongly supported this hypothesis, there have been some conflicting data arguing

against the involvement of FRG1 in FSHD, or at least minimizing its role as a primary

pathogenic insult 32,34,35,38-42,47,49,50,52,53,55,56,146,148,181. Thus, it is fair to say that FRG1 is a controversial FSHD candidate gene 11. Nevertheless, for this study, we were

unconcerned with this ongoing debate, because our primary goal was to demonstrate

proof-of-principle for RNAi therapy of dominant myopathies in general, and the FRG1- high line was useful as an outstanding model of dominant muscle disease 35,39. We reasoned that its involvement in FSHD, or lack thereof, was irrelevant to the goal of this study. We therefore developed a gene therapy strategy to knockdown pathological levels of human FRG1 in FRG1-high mouse muscles. Here, we reported that AAV6-delivered artificial microRNAs reduced toxic FRG1 levels and improved histological and functional muscle abnormalities associated with FRG1 over-expression in mice. Our work therefore supports the therapeutic potential of RNAi therapy for dominant myopathies in general. In addition, it could be applied to FSHD, if additional evidence supporting FRG1 involvement in the disease emerges; alternatively, our strategy could be modified to target other FSHD candidate genes, such as DUX4 32,52,53,55,56,160.

Finally, we note that over-expression of an otherwise normal gene, such as in

FSHD and FRG1-high mice, is a unique pathogenesis mechanism for dominant muscle

diseases. Indeed, most other dominant myopathies arise from point mutations in one

allele of a disease gene, while the other allele remains normal. In some cases, the

remaining normal alleles encode essential proteins, and sufficient levels of the wild-type

allele may be required to maintain some level of normal muscle function. Thus, any

87 therapeutic benefits of reducing the dominant mutant allele could be counterbalanced if a

similar reduction of the remaining wild-type allele causes haploinsufficiency-related myopathy. For example, dominant negative caveolin-3 (CAV3) mutations that result in

97% loss of normal CAV3, cause severe limb-girdle muscular dystrophy type 1C

(LGMD1C), but mutations resulting in 16% or 50% normal CAV3 levels produce only mild hyperCKemia without muscle weakness, or normal phenotypes, respectively

70,128,129. Thus, in addition to dominant mutations, CAV3 loss of function below a certain

threshold also contributes to myopathic phenotypes. This was not a concern in our proof- of-principle study here, because dominant myopathy in FRG1-high mice was caused by

increased dosage of an otherwise normal gene; we therefore only needed to reduce FRG1

to sufficiently non-toxic levels. In contrast, for most other dominant myopathies, such as

the CAV3 example above, disease allele-specific RNAi strategies may be required.

Importantly, several studies support the feasibility of engineering inhibitory RNA

sequences that can distinguish between two alleles differing by a single nucleotide 134.

Each allele-discriminating miRNA must be uniquely designed and empirically validated, since mismatches do not necessarily prevent gene silencing (Fig. 4.1 and 134). Such

strategies often require designing additional mismatches in the miRNA with the goal of

destabilizing interactions with the normal allele of a disease gene 134. Thus, our RNAi

strategy can be modified for disease allele-specificity, when applicable. Our work

therefore supports that RNAi-based gene therapy is a promising candidate strategy for

treating dominant myopathies, regardless of the causal genetic mutation. Future studies

demonstrating the practicability of allele-specific silencing of dominant myopathy genes will further strengthen this conclusion.

88 Figure 4.1. miFRG1 sequences. (a) Sequence and structure of miFRG1 transcripts screened in Figure 1. All miRNAs are derived from human mir-30a structure. Underline sequences indicate the predicted mature guide strand. (b) Position of cognate binding sites in the human FRG1 gene for each miFRG1 sequence. Red bases indicate binding sites. FRG1 translation begins with the ATG at position 192. (c) The binding site for our lead miFRG1 sequence (mi948) has one mismatch with the homologous sequence in the mouse FRG1 gene, but (d) Taqman real time PCR analysis demonstrated that this mismatch was not sufficient to prevent miFRG1-mediated silencing of the endogenous mouse transcript. Data represent means ± s.e.m. using 7 male mice per group. *, significant difference from AAV.miGFP-treated animals, p<0.05 (t-test). Mouse FRG1 levels were normalized to endogenous mouse GAPDH and relative gene expression determined by the D∆Ct method.

89

Continued

90 Figure 4.1 Continued

Continued

91 Figure 4.1 Continued

Figure 4.1

92 Figure 4.2. In vitro screen to identify lead miFRG1 sequences. (a) Plasmids used for testing FRG1 gene silencing in HEK293 cells. (Left) Each FRG1-targeted miRNA

(miFRG1) was cloned downstream of the mouse U6 promoter (U6 pro). U6.miFRG1 expression plasmids were co-transfected with FRG1 target plasmids expressing Renilla luciferase (RenLuc)-FRG1 fusion transcripts (right, top) or the human FRG1 open reading frame (right, bottom). In the Luciferase-FRG1 expression plasmid, human FRG1 was placed downstream of the Renilla luciferase stop codon, thereby serving as a 3’

UTR. This plasmid also contained a separate Firefly luciferase reporter, which was useful as a transfection control. SV40, indicates SV40 promoter; TK, indicates herpes simplex virus (HSV) thymidine kinase promoter. CMV, indicates cytomegalovirus promoter. (b) Luciferase assay screen. FRG1 gene silencing was initially determined by measuring the ratio of Renilla:Firefly luciferase activity from co-transfected cell lysates.

Numbers on x-axis indicate miFRG1 sequences; numbers correspond to position on the

FRG1 cDNA. MiGFP and miLUC control miRNAs do not target FRG1. (c) Relative

FRG1 mRNA and (d) protein expression in HEK293 cells co-transfected with

CMV.FRG1 and indicated miFRG1 expression plasmids. In c, FRG1 levels were determined by Taqman assay and normalized to human β-actin expression. Data represent s.e.m. from two independent experiments performed in triplicate. Western blot in d shows representative data from 3 independent experiments. The U6.miFRG1.948 sequence consistently knocked down human FRG1 levels in all three assays. It was subsequently used in all in vivo experiments.

93

Figure 4.2

94 Figure 4.3. FRG1 gene silencing improves muscle mass in FRG1-high mice. (a) AAV

vectors used for in vivo studies. All contain a CMV promoter-driven humanized Renilla

GFP (hrGFP) cassette with an SV40 polyA (pA) signal. The AAV.miFRG1 and

AAV.miGFP vectors also contain upstream U6.miRNA expression sequences. The miGFP sequences target eGFP, and do not impact levels of hrGFP, which is a different gene. Flanking black rectangles indicate AAV inverted terminal repeats (ITRs). (b) GFP epifluorescence shows near saturation of lower limb musculature in adult mice injected intramuscularly as newborns with 5 X 1010 DNAse Resistant Particles (DRP) of AAV6.

(c) In all groups except FRG1-high mice injected with the AAV.miGFP virus, this delivery method produced high transduction in adductors (add), gastrocnemius (gas), tibialis anterior (TA), and gluteus (glut) muscles, shown by hrGFP epifluorescence in isolated muscles. Transduction was lower and less consistent in the quadriceps (quad).

In contrast, 14 weeks after AAV.miGFP injections, hrGFP epifluorescence was not evident in lower limbs of FRG1-high mice. (d-e) AAV.miFRG1 significantly improved

muscle mass in FRG1-high mice compared to uninjected or control-injected muscles. In d,

vectors were unilaterally injected as indicated. Dorsal (bottom) and ventral (top) views

from the same representative animal are shown. In e, data represent the mean weights ±

s.e.m. of male FRG1-high (+) or WT (-) muscles injected with indicated vectors (n=14

muscles per group). *, indicates significant difference from WT counterpart (ANOVA

with Bonferroni post-test, p<0.05).

95

Figure 4.3

96

Figure 4.4. AAV.miGFP transduction in 3 week-old FRG1-high mice intramuscularly injected with 5 X 1010 DRP AAV.miGFP at post-natal day 1 (P1).

These results confirmed that transduction occurred in AAV.miGFP-injected FRG1-high muscles, and thereby further supported that this non-protective vector was lost over time

(by 14 weeks; Figure 4) with turnover of FRG1-high muscles. In contrast, the therapeutic

AAV.miFRG1 vector remained highly expressed in FRG1-high mice 14 weeks later

(Figures 4.3,4.7).

97 Figure 4.5. HrGFP transduction in isolated muscles from male mice used in this

study. The numbers within each photomicrograph indicate ear tags of individual animals. Numbers located below each muscle are individual weights (in mg).

98 Continued Figure 4.5

99 Figure 4.5 Continued

Figure 4.5

100

Figure 4.6. In vivo knockdown of FRG1 in FRG1-high mice. (a) Taqman assay showed

that AAV.miFRG1 vectors reduced FRG1 mRNA expression by a statistically significant

average of 55% in FRG1-high muscles, compared with AAV.miGFP-injected controls

(n=7; ANOVA). FRG1 expression was determined by Taqman assay (human FRG1 primer/probe) normalized to mouse GAPDH expression. (b) Representative western blot confirmed FRG1 protein knockdown in vivo. F, indicates FRG1-high muscles; (-) indicates uninjected. Tubulin served as a loading control. (c) Immunofluorescence stain of AAV.miFRG1-injected FRG1-high muscle cryosections. Green shows transduced myofibers (hrGFP epifluorescence) and red stain shows FRG1 expression by immunostaining with DMA-AP-1 FRG1 primary antibody and Alexa-594 labeled

secondary antibody. Importantly, FRG1 immunofluorescence inversely correlated with

AAV.miFRG1 transduction, thereby demonstrating the efficacy of our FRG1 knockdown

strategy.

101 Figure 4.7. FRG1 gene silencing improved myopathic histology in FRG1-high mice.

FRG1-high mice show several histological indicators of myopathy, including fibrosis, fat deposition, myofiber size variability, and myofibers with centrally located nuclei. (a)

Control AAV.miGFP vectors had no impact on any of these histolopathological

phenotypes. In contrast, AAV.miFRG1-transduced FRG1-high muscles were

indistinguishable from all WT groups. Muscle cryosections were stained with (left to

right): hematoxylin and eosin (H&E), hrGFP (epifluorescence), FRG1 antibodies

followed by red-labeled secondaries, and DAPI. (b) Oil Red O stain (with corresponding

H&E-stained serial sections) shows fat-infiltrated lesions in AAV.miGFP- but not

AAV.miFRG1-transduced FRG1-high muscles. (c-d) FRG1 gene silencing normalized myofiber size defects and the number of centrally nucleated myofibers in FRG1-high

mice. In c, * indicates significant differences between the comparable WT group,

p<0.05. # indicates significant differences from FRG1 mice injected with miFRG1,

p<0.05. Statistics were determined using one way ANOVA with Kruskal-Wallis post-

test.

102

Figure 4.7 103

Figure 4.8. FRG1 gene silencing improved strength in FRG1-high mice. Hind limb grip strength assay showed that AAV.miFRG1-treated FRG1-high animals were significantly stronger than counterparts injected with AAV.miGFP controls. In contrast, the former group and all injected wild-type controls were not significantly different from uninjected wild-type animals at any time point between 6-10 weeks post-injection. Data represent means ± s.e.m. using 8 male mice per group (ANOVA with Bonferroni post-

test).

104

Figure 4.9. Additional controls for the grip strength assay (Figure 4.8). The

AAV.hrGFP vector contained no U6.miRNA sequences, and injected animals showed similar patterns of muscle strength/weakness that we saw in AAV.miGFP-injected animals. Specifically, AAV.hrGFP-injected and uninjected WT male mice were indistinguishable in grip strength, while AAV.hrGFP-injected FRG1-high animals were

significantly weaker than WT groups.

105

CHAPTER 5

RNA INTERFERENCE INHIBITS DUX4-INDUCED MUSCLE TOXICITY IN

VIVO: IMPLICATIONS FOR A TARGETED FSHD THERAPY

5.1 Introduction

FSHD is often cited as the third most common muscular dystrophy, affecting 1 in

20,000 individuals 150. However, recent data indicate FSHD is the most prevalent

muscular dystrophy in Europe (1 in 7,500), suggesting its worldwide incidence may be

underestimated 1. FSHD is an autosomal dominant disorder that typically arises in young

adulthood, with most patients showing clinical features before age 30 150. Age-of-onset and clinical severity vary, but typically all FSHD patients develop progressive and asymmetrical wasting of facial and shoulder girdle muscles. In addition, ~50% of individuals with FSHD display lower limb weakness and ~20% have abdominal pelvic and girdle muscle involvement, which almost invariably leads to hyperlordosis and wheelchair dependence 150.

There is currently no treatment for FSHD, and despite its relative abundance

among the muscular dystrophies, very few FSHD-targeted translational studies have been

published 203. This historical lack of focus on therapy development was not borne of

neglect, but instead arose because the disease is complex and deciphering its pathogenic

106 mechanisms required nearly two decades of study 20,27,28,32,35,36,51-53,56,154,204,205. Although

the story is likely not complete, the mechanisms underlying FSHD have started clarifying

in recent years, and a new pathogenesis model has emerged 27,32. Unlike most typical

Mendelian disorders, FSHD is not caused by mutations in the coding region of a single

gene. Instead, FSHD requires complex genetic and epigenetic changes that conspire to

permit expression of one or more myopathic genes 20,32,35,36,154. Several FSHD candidate

genes have been identified, but numerous recent studies support that the primary

contributor to FSHD pathogenesis is the pro-apoptotic DUX4 gene, which encodes a

transcription factor 27,32,51-53,55,56. Thus, in the simplest terms, DUX4-overexpression is a

primary pathogenic insult underlying FSHD.

The emergence of DUX4 as an important prospective therapeutic target has

therefore lowered the barrier to pursuing translational research for FSHD. We

hypothesized that reducing DUX4 expression using RNA interference (RNAi) would

offer a potential treatment for the disease. In this study, we provide evidence that DUX4

gene silencing, triggered by engineered artificial microRNAs, is myoprotective in vivo.

Our data demonstrate proof-of-principle for DUX4 gene silencing as a promising

therapeutic approach for FSHD.

5.2 Materials and Methods

Cloning of DUX4-targeted microRNAs. Five different mouse U6 promoter-driven

artificial microRNAs targeting human DUX4 (called miDUX4s) were cloned as previously described206,207. Each U6.miDUX4 construct was derived from human mir-30

stem and loop sequences and structures, but the 22 nucleotide (nt) mature mir-30 duplex

107 was replaced by sequences targeting the DUX4 gene (Fig 5.1). The miDUX4

nomenclature refers to the first position of the miRNA binding site relative to the +1 of

the DUX4 coding region (MAL start). The control U6-driven artificial miRNA targeting

eGFP (miGFP) was previously described 207.

Luciferase assay. The luciferase reporter plasmid (Fig 5.2 a) was modified from

Psicheck2 (Promega, Madison, WI) 206. Human DUX4 cDNA was cloned downstream of

the Renilla luciferase stop codon, thereby functioning as a 3’ UTR. A separate TK

promoter driven Firefly luciferase cassette, present on the same plasmid, served as a

transfection control. HEK293 cells were co-transfected in triplicate wells

(Lipofectamine-2000, Invitrogen, Carlsbad, CA) with the luciferase DUX4 reporter and

individual U6.microRNA expression plasmids in a 1:5 molar ratio. DUX4 gene silencing

was determined by measuring Renilla and Firefly luciferase activity (Dual Luciferase

Reporter Assay System, Promega) 24 hours post-transfection, following manufacturer’s

instructions. Triplicate data were averaged, and individual experiments performed three

times; results were reported as the mean ratio of Renilla to Firefly activity ± SEM.

Western blot. For in vitro work, HEK293 cells were co-transfected with U6.miDUX4 or

control microRNA plasmids and a CMV.DUX4 expression vector at an 8:1 molar ratio.

Protein was extracted 48 hours later (M-PER from Pierce, Rockford. IL). For in vivo

work, protein was extracted from muscles injected 1.5-4 weeks prior, using previously

described methods. Protein was quantified by Lowry assay (BioRad, Hercules, CA), 30

µg samples were separated on 12% SDS-PAGE, transferred to PVDF membrane, and incubated with the following antibodies: mouse monoclonal antibody to V5 (HRP- coupled) (1:5,000 Invitrogen, Carlsbad, CA); mouse monoclonal GAPDH antibody

108 (1:500, Millipore, Billerica, MA) overnight at 4°C. GAPDH-probed blots were washed,

then incubated with HRP-coupled goat anti-mouse secondary antibody (1:100,000;

Jackson ImmunoResearch, West Grove, PA) for 1 hr at room temperature. Following

washes, blots were developed using Immobilin Western HRP substrate (Millipore,

Billerica, MA), and exposed to film.

AAV vector delivery to mouse muscle. U6.miDUX4.405 was cloned into

AAV.CMV.eGFP (aka AAV.eGFP) proviral plasmid upstream of CMV.eGFP.

AAV6.DUX4 (aka AAV.CMV.DUX4) was previously described 56. AAV6 particles were generated and titrated as previously described by the Viral Vector Core Facility at

The Research Institute at Nationwide Children’s Hospital. Eight-week-old C57BL/6 female mice received 50µl direct intramuscular injections into the tibialis anterior (TA).

For all experiments except grip strength, pre-mixed virus cocktails contained 8 x 108

DNAse resistant particles (DRP) of AAV6.DUX4 and 3 x 1010 of either AAV6.miDUX4

or AAV6.eGFP. For grip strength studies, animals received 3 x 109 DRP of

AAV6.DUX4. All mouse procedures were performed following guidelines approved by

the Institutional Animal Care and Use Committee (IACUC) at the Research Institute at

Nationwide Children’s Hospital.

Imaging and histology. In vivo AAV transduction was determined by eGFP epifluorescence using a fluorescent dissecting microscope (MZ16FA, Leica, Wetzlar,

Germany) at x4.63 magnification. Dissected muscles were placed in O.C.T. Compound

(Tissue-Tek, Torrance, CA), frozen in on liquid nitrogen-cooled isopentane, cut onto slides as 10 µm cryosections, and stained with hematoxylin and eosin (H&E; following standard protocols)56, V5 (Millipore, Billerica, MA ) and cleaved Caspase-3 (Cell 109 Signaling Technology, Inc., Danvers, MA) polyclonal antibodies. For

immunohistochemistry, cryosections were post-fixed in 4% paraformaldehyde, washed, blocked in PBS with 5% goat serum and 0.3% Triton X-100, incubated overnight at 4°C with primary antibody (cleaved Caspase-3 1:1500 and V5 1:1000; in 1% BSA, 0.3%

Triton X-100, and PBS), and then with AlexaFluor-594 conjugated goat anti-rabbit secondary antibodies (1:500; 1 hr at RT; Molecular Probes, Carlsbad, CA). Slides were covered in Vectashield plus DAPI. Muscle cross-sectional fiber diameters and percentage of myofibers with centrally-located nuclei were determined from muscles injected 3 and 4 weeks prior (n = 4 muscles per group; 5 representative 20X photomicrographs per section), using cellSens Version 1.3 software (Olympus

Corporation, Center Valley, PA).

Real-Time Polymerase Chain Reaction. RNA was extracted from muscles 2-3 weeks post injection (N=7 muscles per treatment; TRI Reagent, Molecular Research Center Inc.

Cincinnati, OH; followed by RNeasy mini column purification, Qiagen, Valencia, CA).

RNA was quantified (NanoDrop ND-1000 Spectrophotometer, Thermo Scientific,

Wilmington, DE) and DNase treated for 30 minutes at 37°C (DNA-free, Ambion, Foster

City, CA). Following DNAse inactivation, RNA was reverse transcribed using random hexamers (High Capacity cDNA Reverse Transcription Kit, Applied Biosystems, Foster

City, CA) and gene specific primers to the DUX4 coding region (5’-

GCTAGCCACCATGGCCCTCCCGACAC-3’) and the DUX4 coding region with an additional tag sequence (5’-

CGACTGGAGCACGAGGACACTGACGATGCCCGGGTACGGGTTCCGCTCAAAG

C-3’)51. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed 110 using a SYBR green reaction (SA Biosciences, Frederick, MD) for DUX4 (5’-

GCTAGCCACCATGGCCCTCCCGACAC-3’ and 5’-

CGACTGGAGCACGAGGACACTGA-3’) and GAPDH (5’-

CACGGCAAATTCAACGGCACAGTCAAGG-3’ and 5’-

GTTCACACCCATCACAAACATGG-3’)51. Caspase-3 levels were detected using a

Taqman Assay primer/probe set (Applied Biosystems, Foster City, CA) and normalized

to GAPDH. All samples were run in triplicate.

Grip strength. Grip strength was measured in forelimbs and hindlimbs of C57BL/6

mice one week prior to injection to establish a baseline, and then weekly up to 4 weeks

post-injection as previously described (n=7 animals per group) 56. Data represent mean

of hindlimb force divided by forelimb force ± SEM.

5.3 Results

Cellular RNA interference pathways can be co-opted to suppress dominant

disease genes for therapeutic purposes 98,105,200,207,208. The goal of this study was to

develop a prospective treatment for FSHD using a DUX4-targeted, RNAi-based gene

silencing approach. To do this, we engineered five different, mir-30-based artificial microRNAs (miDUX4s; Fig 5.1) targeting the human DUX4 open reading frame. We cloned each miDUX4 sequence into an U6-promoter-driven expression cassette, and performed two in vitro screening assays to identify our lead DUX4-targeted microRNA

(Fig 5.2)206,207. First, we co-transfected each U6.miDUX4 construct, or a control miRNA

vector, with a dual luciferase plasmid in which DUX4 was cloned as the 3’ UTR of

Renilla luciferase. A separate Firefly luciferase cassette was used as an internal

111 transfection control (Fig 5.2 a). We then determined the degree of DUX4 gene silencing

by measuring the ratio of Renilla to Firefly luciferase activity 24 hours later (Fig. 5.2 a).

Two constructs (miDUX4.405 and miDUX4.1156) consistently reduced Renilla.DUX4 activity (Fig. 5.2 b). We next confirmed their ability to silence DUX4 at the protein level

(Fig. 5.2 c), and found that miDUX4.405 was consistently superior in vitro. We therefore

selected miDUX4.405 as our lead sequence for subsequent in vivo experiments (and heretofore refer to the 405 sequence as miDUX4).

For in vivo gene delivery, we cloned U6.miDUX4 into an adeno-associated viral vector (AAV) separately expressing a CMV.eGFP reporter cassette, and generated AAV6 particles (Fig. 5.3 a). No stably expressing DUX4 animal model exists. As a result, we relied upon our previously published AAV-based DUX4 mouse model to express myotoxic levels of DUX4 in wild-type mouse muscles 56. Importantly, this model rapidly develops dose-dependent myopathic phenotypes consistent with those reported in other models of muscular dystrophy, including myofiber degeneration and regeneration, fibrosis, muscle weakness, and elevation of cell death pathways 56.

To assess the effects of miDUX4 on DUX4-induced myopathy in vivo, we directly

injected wild type C57BL/6 tibialis anterior (TA) muscles with an AAV cocktail

containing AAV6.DUX4 with either AAV6.miDUX4 or the control vector (CMV.eGFP)

(Fig. 5.3 a). Control-treated DUX4-expressing muscles showed histological evidence of

muscle degeneration and regeneration, 2, 3 and 4 weeks after injection (Fig. 5.3 b). In

contrast, AAV.miDUX4 protected co-injected DUX4-expressing muscles from

degeneration (Fig. 5.3 b). Muscles from this group of animals were histologically normal

at all time points examined (Fig. 5.3 b). In addition to obvious visual differences in

112 miDUX4- and control-treated muscles, significant quantifiable changes were evident

(Fig. 5.3 c-d). Specifically, muscles injected with AAV6.DUX4 alone or AAV6.DUX4

and AAV6.eGFP vectors contained abundant myofibers with reduced size and centrally-

located nuclei, which are both indications of muscle degeneration and regeneration

typically associated with myopathy. In contrast, myofiber sizes and centrally-located

nuclei were normal in muscles co-injected with AAV6.DUX4 and AAV6.miDUX4

vectors (Fig. 5.3 b-d). Improved muscle histology in miDUX4-treated muscles correlated with 90% and 64% average reduction in DUX4 protein and mRNA, respectively, compared to uninhibited DUX4 expression in controls (Fig. 5.4 a-b).

We next determined if AAV.miDUX4 protected muscles from pathological molecular changes associated with FSHD that were downstream of DUX4. For this, we focused on caspase-3, which is expressed in myofibers of FSHD patients and activated by

AAV6.DUX4 expression in mouse muscle 46,56,162. Consistent with our previous results,

we found that uninhibited DUX4 expression was associated with caspase-3 positive

lesions in AAV6.DUX4-transduced control muscles (Fig. 5.5 a). In contrast, there were

no caspase-3 positive myofibers in muscles co-injected with AAV6.DUX4 and

AAV.miDUX4 vectors (Fig. 5.5 b), and caspase-3 transcripts were reduced by 77% in

AAV.miDUX4-treated animals (Fig. 5.5 c).

Finally, we measured the effects of AAV6.miDUX4 on DUX4-associated hind

limb grip strength deficits in mice. To do this, we measured forelimb (uninjected) and

hindlimb (injected) bilateral grip strengths weekly in wild-type mice following injection

with saline or the indicated vectors (Fig. 5.6). By two weeks, mice injected with

AAV6.DUX4 alone or AAV6.DUX4 with control AAV.eGFP showed significantly

113 reduced grip strength compared to all other groups. This timepoint is consistent with the onset of degeneration in muscle cryosections (Figs 5.3 and 5.6). Weakness resolved by 3 weeks, as regenerative processes were underway (Figs 5.3 and 5.6). In contrast, animals co-injected with AAV6.DUX4 and AAV6.miDUX4 were not significantly weaker than saline-injected wild-type mice at any timepoint following injection. Mice that received

AAV6.miDUX4 alone were unaffected, suggesting that miDUX4 expression was well- tolerated by normal muscles in the absence of a gene target. Together, these results provide evidence for DUX4 inhibition as a therapeutic strategy for FSHD.

5.4 Discussion

Developing treatments for genetic diseases has long been a major goal of biomedical research. Over the years, numerous lethal or debilitating disorders have been transformed into manageable ailments. To date, the list of inherited diseases for which effective treatments exist does not include the muscular dystrophies. Still, much progress has been made in the muscular dystrophy field in the last 3 decades, beginning with the landmark discovery that DMD gene mutations were responsible for Duchenne muscular dystrophy (DMD) in 1986 209. Today, because of similar gene mapping efforts, mutations in at least 40 known genes have been linked to various types of muscular dystrophy. Typically, these disorders arise from mutations that alter the expression, structure, and/or function of a single protein. Identifying a monogenic, protein-coding mutation is an optimal outcome for a gene mapping study, because a relatively straightforward path can be followed toward cell and animal model development, which then provides a platform for testing hypotheses about disease pathogenesis and ultimately

114 developing therapies. For example, DMD gene discovery led to the 1988-89

identification of the mdx mouse as a DMD model 210,211, which has been used extensively

to study DMD pathogenesis and test therapeutic strategies, some of which are now in the

clinical pipeline as prospective DMD treatments 212. This progress would have been

impossible without knowledge of the DMD gene mutation or the mdx mouse model.

In contrast to DMD, FSHD translational research has been hindered for decades

by uncertainty about disease pathogenesis and the absence of an animal model. As a

result, the comparatively few FSHD therapy studies published to date have primarily focused on palliative care, surgical interventions, or treatments that addressed some non-

specific myopathic features common to many types of muscular dystrophy 203. However,

the emergence of recent data supporting DUX4 de-repression as a key event in FSHD

pathogenesis suggests the field has passed a threshold of understanding, making FSHD-

targeted translational research more feasible. Despite the recent turn of events, barriers to

translation remain. Arguably the most important such barrier is that a “traditional” FSHD

animal model (e.g. one expressing a heritable DUX4 transgene) is still lacking.

Following a conservative strategy toward therapeutic development would necessitate

delaying in vivo testing of DUX4-targeted approaches until such a model was available.

However, this delay would only add to an already potentially long translational path.

Indeed, the mdx mouse has been available for nearly 25 years and a DMD “cure” is yet

unrealized, so even if a transgenic or knock-in DUX4-expressing FSHD model were in

hand today, rapid development of an FSHD therapy would not be guaranteed. Thus, to

avoid further delay, we began testing our DUX4 gene silencing approach in vivo using an

115 alternative, AAV-based model of DUX4-associated myopathy, which we previously published 56.

In our AAV model, DUX4 is expressed at very high levels (using a CMV promoter and myotropic AAV6 capsid) in essentially every myonucleus within a transduced region. The expression levels produced are likely much higher than DUX4 levels detected in biopsies from FSHD patients 51,52. Indeed, DUX4 so far appears relatively rare in human muscle. One study estimated its presence in only 1 in 1,000 –

10,000 FSHD myonuclei, and up to 70 cycles of PCR were required to detect the endogenous transcript 51. This apparent scarcity could argue against the involvement of

DUX4 in FSHD pathogenesis. Specifically, how could such a rare transcript/protein cause the devastating clinical phenotype sometimes seen in some FSHD patients? This question still remains unresolved in the field. However, one plausible explanation is that a tolerable steady state of DUX4 is permissible in human muscle, and a stochastic series of FSHD-associated DUX4 amplifications produce small, myopathic events that cumulatively lead to significant muscle loss over time. Thus, observing a rare DUX4 amplification event in a small muscle biopsy sample would be difficult at any single time-point.

The dose dependency of DUX4 toxicity lends support to this amplification model.

Indeed, in a previous study we found that high doses of AAV6.DUX4 produced widespread lesions throughout an entire mouse TA muscle, but damage manifested more slowly and focally at decreasing dosages 56. In this study, DUX4 gene silencing by

AAV.miDUX4 protected muscles from DUX4-induced damage, even though the protein was still detectable in virtually all transduced myonuclei (Fig. 5.5). In addition, our in

116 vivo work here is consistent with a recently published study showing DUX4-targeted

siRNA or antisense RNA could inhibit an atrophic phenotype in cell culture 213.

Together, these studies support the feasibility of DUX4 inhibition by RNAi as a potential

therapy for FSHD.

Before moving these pre-clinical studies forward, issues common to other nucleic acid-based therapies must be addressed. One is safety. RNAi therapy is still a new and clinically unproven approach. At very high doses, exogenous inhibitory RNAs may sometimes saturate natural miRNA biogenesis pathways, thereby potentially causing non- specific cytotoxicity. Sequence-specific off-target effects can also occur if an inhibitory

RNA shares complementary sequence with other, unintended transcripts 100,105.

However, over-expression related toxicity was associated with sub-optimally processed,

first generation shRNA vectors, while the miRNA-based approach we used here

generates a transcript that cycles through the miRNA biogenesis pathway efficiently and

thus avoids this toxicity 105. Importantly, our proof-of-principle studies here already support the safety of our approach, as high doses of AAV6.miDUX4 vector produced no overt functional pathology in wild-type mice lacking a DUX4 target (Fig. 5.6).

Moreover, bioinformatics analyses support that our miDUX4 sequences are DUX4-

specific and should not directly silence off-target transcripts. Additional dose escalation studies and off-targeting analyses in mice, human cells, and perhaps non-human primates, will further solidify the safety of our miDUX4 vectors.

A second major issue that our strategy shares with other nucleic acid therapies is delivery. Currently, AAV vectors are the tool of choice for muscle gene delivery because of their myotropism and excellent safety profile. Indeed, AAV has already proven safe

117 and effective in human clinical trials 199,214, supporting that our approach potentially has

immediate practicability if myofibers are the only DUX4-expressing target cells.

However, it is difficult to predict the effectiveness of our AAV.miDUX4 approach should

muscle progenitor cells also require DUX4 inhibition. This uncertainty arises from the

lack of data about AAV tropism in muscle satellite cells (SC). It will therefore be

important to determine if FSHD pathology is related to DUX4 expression in SCs, and if

so, whether SC transduction is possible with AAV6 or another AAV serotype. If DUX4

suppression in SCs is required, and AAV vectors are non-transducing, it may be possible

to rationally design or pan for SC-tropic AAV capsids to enhance efficacy 215,216.

Alternatively, in vitro-synthesized inhibitory RNAs could be used as cell-penetrating delivery strategies continue to improve.

A final potential complicating issue related to a DUX4-targeted therapy is the necessity for allele-specific silencing. Specifically, DUX4 can be alternatively spliced to produce a short, apparently non-toxic form lacking the C-terminal transactivation domain

(DUX4-s)160. Although the function of DUX4-s is uncertain, and it too seems to be

expressed at similarly low levels (like full-length DUX4; DUX4-fl) in muscle, there is a

hypothesis currently in the field that DUX4-s contributes some important function to normal muscle, while preferential expression of the full-length DUX4 isoform in FSHD contributes to myopathic phenotypes. If true, it may therefore be necessary to specifically silence only DUX4-fl. Importantly, some of our miDUX4 constructs can accomplish this allele-specificity, including mi1156 (Fig. 5.2 b-c).

In conclusion, we have entered an exciting period in FSHD research. New basic discoveries on FSHD pathogenesis have opened the door to begin developing, arguably

118 for the first time, FSHD-specific therapeutic strategies. Such approaches will likely include DUX4 inhibition, as we described here using RNAi. Nevertheless, although the pathogenic mechanisms underlying FSHD are clarifying, the picture is likely still incomplete. Our current and previous studies 207 set the stage for applying RNAi more broadly to other FSHD-associated gene targets, as they emerge.

119

Figure 5.1. DUX4 microRNAs (a) miDUX4 sequences. Mature DUX4-targeted guide strand (red) and passenger strand (blue) are shown for each miDUX4 sequence.

Arrowheads show Drosha (grey) and Dicer (black) cut sites. (b) DUX4 cDNA beginning at the first homeodomain (position 281). Red nucleotides denote binding sites for miDUX4 sequences.

120 Figure 5.2. In vitro screen of miDUX4 sequences. (a) Plasmids used to test DUX4 gene silencing in HEK293 cells. For luciferase assay, the DUX4 cDNA sequences were cloned downstream of the Renilla luciferase (Ren Luc) stop codon, thereby functioning as a 3’

UTR. The Ren Luc.DUX4 fusion transcript was driven by the SV40 promoter. A separate Firefly luciferase cassette was transcribed by the herpes simplex virus thymidine kinase (TK) promoter. CMV indicates cytomegalovirus promoter. (b) Luciferase assay screen to identify lead miDUX4 constructs. DUX4 gene silencing was determined by measuring the ratio of Renilla to Firefly luciferase activity from cells co-transfected with the reporter plasmid and U6 promoter (U6 pro)-driven miDUX4 sequences. (c) Western blot to confirm gene silencing efficacy of miDUX4.405 and miDUX4.1156 following transfection of indicated U6.miRNAs and CMV.DUX4 in HEK293 cells. GAPDH serves as a loading control. UNT indicates untransfected HEK293 cell extracts.

121

Figure 5.2

122 Figure 5.3. DUX4-targeted artificial miRNAs improve DUX4-associated myopathy

in vivo. (a) Adeno-associated (AAV) vectors used for in vivo studies. Black boxes

indicate AAV inverted terminal repeats (ITRs) and pA indicates an SV40

polyadenylation signal. (b) By 2 weeks, control-treated, DUX4-expressing muscles (left) show histopathological evidence of degeneration including: myofibers with central nuclei, inflammatory invasion and fibrosis. miDUX4-treated muscles (right) are normal at all timepoints. Scale bar = 50 µm. (c-d) Histological changes were quantified at 3 and

4 weeks post injection. (c) shows the percentage of myofibers containing centrally located nuclei (C. N.), an indication of muscle regeneration following damage, 3- and 4-

weeks post-injection. (d) myofiber size distributions in indicated groups. A shift toward

small-bore myofibers is characteristic of myopathy.

123

Figure 5.3

124

Figure 5.4. DUX4 knockdown in vivo. (a) Western blots demonstrate DUX4 protein knockdown at 90% in miDUX4-treated muscles compared to control-treated muscles.

DUX4 silencing was evident at all time points examined, but only the 2- and 3-week samples are shown here. GAPDH served as a loading control. (b) Real-time PCR shows miDUX4 reduced DUX4 mRNA by 64% in vivo (n=7 muscles per group from combined

2 and 3-week groups). * indicates significant difference from AAV.eGFP injected group, p<0.05 (ANOVA).

125 Figure 5.5. miDUX4-treated myofibers are caspase-3 negative. (a) Serial sections of eGFP control treated muscle mimics immunohistochemical staining patterns of DUX4- induced myotoxicity indicated by various symbols. Asterisks represent healthy fibers that have DUX4 (+) nuclei but are cleaved caspase-3 (-)/GFP(+) . Arrows and arrow heads point to degenerating myofibers: arrows DUX4 (+)/cleaved caspase-3 (+)/GFP (-) and arrow heads DUX4(+)/cleaved caspase-3(+)/GFP(+). (b) Treatment with miDUX4 protects myofibers from degeneration indicated by the absence of cleaved caspase-3 (+) myofibers and very few myofibers with DUX4 (+) cytoplasmic staining. Scale bar =

50µm. (c) Relative caspase-3 expression of miDUX4 and eGFP treatment groups as determined by q-PCR.

126

Figure 5.5

127

Figure 5.6. AAV.miDUX4 protects mice from DUX4-associated grip strength deficits. Mice injected with DUX4 and DUX4 with control eGFP show significant hindlimb strength deficits compared to all controls at 2 weeks post injection. Mice that received DUX4 with miDUX4 treatment were showed no significant difference in strength from controls. Force represents the grip strength ratio of hindlimb (injected) to forelimb (uninjected). N = 8 mice per group ***p<0.001 two-way analysis of variance with Bonferroni post-test.

128

CHAPTER 6

GENERAL DISCUSSION

6.1 Relevance of this work

FSHD is the most prevalent muscular dystrophy with an incidence up to 1 in

7,5001. However, this statistic is minimal compared to the number of people actually affected by the disease. In reality, FSHD affects the patients suffering from the disease as well as caregivers, friends and health care professionals providing physical, emotional, and financial support. At the moment FSHD can be a devastating and frustrating diagnosis to receive because there is no effective treatment. This absence is due to an undefined disease etiology and while great strides have been made in the last couple years, there is still an insufficient understanding. Complications associated with extreme clinical variability and complex genetic criteria have made studying and modeling the disease problematic. The goal of the studies presented here were to help define the pathogenic mechanism of FSHD and develop a potential treatment.

6.2 DUX4 discussion

Just as clinicians had to define standards for FSHD diagnosis4, researchers had to define standards for FSHD candidate gene studies. Since 1994, the FSH Society has

129 annually held the FSHD International Research Consortium Meeting where topics like

that could be addressed. Each year a dynamic set of criteria are presented, debated and

adjusted to reflect the most recent FSHD findings. Based off of these, we considered

four main criteria while performing our FSHD candidate gene studies. Basically a FSHD

candidate gene should (1) be located in the area of the FSHD mutation on chromosome 4;

(2) show consistent over-expression in muscles from FSHD patients; (3) have the capability to damage muscles when over-expressed in vivo, and (4) be activated specifically in FSHD-affected muscles (e.g. facial, shoulder-girdle, limb muscles) and/or non-muscle areas of pathology (retina, inner ear). The first part of the work described here was to explore criterion 3. Chapter 3 discussed the myotoxic capability of DUX4 in vivo and helped define the role of DUX4 in FSHD pathogenesis. We found DUX4 could dose dependently induce myopathic phenotypes in zebrafish and adult wild type mouse muscle when expressed in vivo. We proved this toxicity was specific to DUX4 expression and not just a side effect of over-loading the cells with protein. Our work also

revealed a p53-dependent mechanism for DUX4-induced myopathy.

Since then multiple studies have been published to support the role of DUX4 in

FSHD and indeed it is now considered the top candidate gene32,51,213. I use the word candidate gene however, because there are still many unknown factors. We do know

DUX4 is located in the correct region, its mRNA is necessary for FSHD manifestation, and as just discussed in Chapter 3, DUX4 is damaging to muscle. However the other two

FSHD candidate gene criteria are less defined. DUX4 has been detected from patient samples, but consistent over-expression is a relative term. DUX4 has been extremely difficult to detect. To date, this has been associated with a theory that there is simply a

130 low abundance of DUX4 mRNA present at any given time. One study has detected a rate

of 1 in 1,000 FSHD muscle cell nuclei expressing DUX4 mRNA51. The predicted

pathogenic mechanism to accompany these findings states that the relatively few nuclei

detected express abundant amounts of DUX4 mRNA and protein leading to the death of

those cells and potentially the surrounding cells as well51. Over time, these isolated

incidences would compound to produce the myopathy described in FSHD. Indeed our

data supports that low doses of DUX4 are capable of inducing wide spread cell death, but

other factors could also play into these low statistics. First, the study indicating the

prevalence of DUX4-expressing nuclei were derived from patient biopsies. Just as

treatment varies from clinician to clinician, sample collection also suffers from

standardization. Different techniques and even different muscles could add to

consistency discrepancies. Second, due to the heterogeneous nature of the disease, these

biopsies may have simply missed more affected areas of the muscle. Third, FSHD is a

progressive disease where patients often describe times of symptom stability interrupted

by times of rapid decline217. Perhaps it is only during these periods of rapid deterioration

when DUX4 is expressed at abundant or higher levels. Finally, DUX4 could be

preferentially expressed in satellite cells, which would represent a very low percentage of

a total muscle sample and could potentially require a state of active regeneration55.

Whatever the case, this is still an unexplained part of the DUX4/FSHD pathology.

Another unexplored area of DUX4 research is its expression pattern. Little is

known about when and where DUX4 is expressed normally and under disease specific

conditions. Certainly further investigations need to identify if over-expressed DUX4 is specifically expressed in FHSD affected tissues. The normal expression and function of

131 DUX4 will also be a critical factor in considering treatments like the gene silencing

studies presented in Chapters 4 and 5. In these proof-of-principal studies we demonstrated the robust silencing capabilities of our AAV delivered therapeutic miRNAs to both FRG1 and DUX4 genes respectively. If the normal function of DUX4 were in any way involved in a critical regulatory mechanism, extra precautions would be necessary when considering the safety of a silencing treatment.

6.3 RNAi therapy discussion

Moving forward clinically with any new treatment strategy inevitably presents with both excitement and safety concerns. As discussed in Chapters 2, 4 and 5, RNAi, like any other new pharmaceutical or genetic treatment, could potentially have risks and problems including possible off-target effects, suboptimal dosing and delivery complications. However, the potential benefits for FSHD and other dominant dystrophies are clearly defined. Our results in Chapters 4 and 5 indicate targeted therapeutic miRNAs are capable of protecting muscle from all disease related phenotypes. We showed significant histological improvements in FRG1-treated mice including: increased mass, consistently sized myofibers, and elimination of fibrosis and fat infiltration. Muscle function and strength were also restored making FRG1-treated mice virtually indistinguishable from wild type litter mates. Importantly, the beneficial effect resulted from a single injection that was sufficient to sustain these improvements for the length of the study. In a similar fashion, DUX4 targeted miRNAs also had undeniably positive results. The AAV.DUX4 adult mouse model induces rapid degeneration and subsequent regeneration in 4 weeks time. This effect is dose-

132 dependent, with higher doses inducing more rapid cell death and measurable functional deficiencies. Importantly, miDUX4 protected the mouse muscles from any sign of degeneration at all doses tested in the 4-week period. Together, these results demonstrate

the potential for RNAi as a successful targeted treatment for a large group of myopathies

including FSHD.

One unknown factor that would have large implications concerning RNAi as a

treatment strategy is time of disease intervention. All studies presented here are

prevention studies, whereas patients would likely benefit from a reversal. One study has

delivered therapeutic shRNAs to aged FRG1 mice and indicated reversal benefits, though

these were only slight improvements208. More troubling studies have been performed in other progressive neuromuscular diseases, specifically in the spinal muscular atrophy

(SMA) field. Patients with SMA suffer from reduced levels of the survival motor neuron

(SMN) protein218,219. It is a devastating neurodegenerative disease where loss of motor neurons leads to the progressive atrophy of proximal muscles, respiratory distress and subsequent death. SMA is subdivided into several groupings based on disease onset severity and outcome. There is an inverse relationship of SMN protein and disease

severity, i.e. the more SMN protein a patient produces the less severe the disease220,221.

Logically the best therapeutic strategy would be to increase amounts of SMN in these

patients and due to the unique pathogenic mechanism of SMA there are multiple

strategies and molecular tools to accomplish that goal222-230. However several studies in

various mouse models have indicated there is an early window for disease intervention

and past that critical time point the strategies are ineffective 226-230. Specifically, one

study demonstrated a treatment was capable of extending the survival curve of SMA

133 mice beyond an unprecedented 250 days when treatment was delivered at postnatal day

2227. However when the same treatment was delivered a few days later, benefits were stunted. Finally, a treatment intervention 10 days after birth resulted in no observed benefit, undoubtedly demonstrating the necessity for correct timing in disease intervention227. When considering neurodegenerative diseases like SMA or progressive dystrophies like FSHD, therapeutic delivery clearly may need to come before or at the very beginning of disease onset. Indeed, FSHD patient muscles rarely show signs of significant regeneration and perhaps for RNAi treatments to be effective they would need to stop the cell death before it occurs, as opposed to trying to fix it once the healthy muscle is gone. Just as in SMA, what may be effective early on may have no significant effect if delivered too late in FSHD progression. Another point to consider is the goal of any treatment, which is to improve patients’ quality of life. Perhaps in that light disease prevention is the best treatment option? Whatever the case, the use of RNAi as a treatment for FSHD and other dystrophies has the potential efficacy, but the timing of therapeutic intervention will require careful examination. For FSHD, this poses yet another obstacle.

Undefined pathogenic mechanisms have hindered therapy development in multiple ways. Other than having an undefined target, as discussed earlier, there is also no stably expressing FSHD animal model. The AAV.DUX4 model presented in chapters

3 and 5 is the best model to date, but it is a transient system. Disease progression is virtually nonexistent because AAV does not integrate into the genome. This means once

DUX4 is expressed in the muscle it immediately induces cell death and the muscle is subsequently repopulated with wild type cells. It is also artificial due to the use of a

134 CMV promoter and a SV40 polyA. While the benefits for proof-of-principle experiments are obvious, a model more faithfully recapitulating the specific FSHD features like the natural promoter and the pLAM polyA may be necessary for evaluating conditions like timing, dose and delivery.

6.4 Concluding remarks

In the preceding chapters we sought to evaluate the role of the FSHD candidate gene, DUX4. We tested the gene’s myopathic potential in vivo, in hopes of better understanding the pathogenic mechanisms responsible for FSHD. Indeed our work supported a causative role for DUX4 in the disease. This kind of study is relevant to affected patients because a defined disease etiology has proven necessary for treatment development. Currently there is no effective treatment for FSHD. In that light we next introduced a possible new treatment option for FSHD and other dominant dystrophies for which RNAi is indicated as a tool for disease specific gene silencing. Our proof-of-

principle study in FRG1 transgenic mice demonstrated the protective affect RNAi

therapy could provide to dominantly inherited myophathies. With that information and

increased evidence of DUX4’s role as the initial pathogenic insult in FSHD, we then

applied the same principles from the FRG1 study to targeting DUX4. All RNAi studies

performed demonstrated significant improvements, clearly defining the method as a

promising therapeutic tool. The goals of this study were met. We successfully

contributed to the understanding of the pathogenic disease mechanisms and developed

targeted translational therapies with the potential to treat FSHD. There is a profound old

saying that simply states knowledge is power. I believe that statement to be the basis of

135 all past, present and future research. In terms of FSHD, there are many basic biological questions that remain unanswered, but as we continue to learn more about DUX4 and the

FSHD pathogenic mechanism we will gain a better understanding of how to treat patients in the most effective way possible.

136

APPENDIX A

RAPID CLONING AND VALIDATION OF MicroRNA SHUTTLE VECTORS: A PRACTICAL GUIDE

A.1. Introduction

RNA interference (RNAi) has emerged as an important modulator of gene

expression in eukaryotic cells69,231. RNAi refers to post-transcriptional control of gene expression mediated by small non-coding microRNAs (miRNAs), which are naturally occurring transcripts encoded in the genomes of a variety of organisms ranging in complexity from single-celled algae to humans70,73,97,232,233. Functionally, miRNAs

reduce protein expression by directing cellular gene silencing machinery to specific target

messenger RNAs (mRNAs), thereby preventing or reducing their translation. Specificity

is accomplished by base-pairing between miRNAs and complementary sequences on

target messenger RNAs (mRNAs)70,73,95,130. Over the last several years, numerous studies have described the cellular pathways controlling miRNA biogenesis and gene silencing70,71,81-83,87,234-242. An important consequence of this growing knowledge is the

development of RNAi as a molecular tool used to investigate basic biological questions

or as a therapeutic243.

A.2. MiRNA Biogenesis

137 Artificial miRNAs, and other small inhibitory RNAs that mimic natural miRNA

structures, can be engineered to potentially silence any gene of interest. It is therefore

helpful to understand natural miRNA biogenesis when developing RNAi as a tool or

therapeutic (Fig. A.1).

In cells, miRNAs are first transcribed from the genome as primary miRNA

transcripts (pri-miRNAs), which form intramolecular hairpin (i.e. stem-loop) structures78,80-84,244. Subsequently, pri-miRNAs undergo a series of processing events,

catalyzed by several proteins, to generate the mature miRNA, which is typically 19-25

nucleotides (nt) long. First, in the nucleus, sequence and structural elements in the pri- miRNA transcript direct a microprocessor complex containing Drosha and DGCR8 proteins to cleave the RNA at a specific location84,87,241. DGCR8 binds the pri-miRNA and serves to correctly position Drosha at base of the stem, where it then makes a staggered cut, producing a shorter (~65-70 nt) pre-miRNA hairpin containing a di- nucleotide 3’ overhang. The nuclear export factor, exportin-5, then binds the 3’ overhang and transports the pre-miRNA through the nuclear pore242. In the cytoplasm, the ribonuclease Dicer then binds the 3’ overhang and cleaves the pre-miRNA two RNA helical turns away (approximately 21 nucleotides)234,235. This cut removes the loop

region and produces the mature duplex miRNA containing di-nucleotide 3’ overhangs at

both ends. One strand (the antisense “guide” strand) is then incorporated into a

riboprotein multimer called the RNA Induced Silencing Complex (RISC)130,240. This miRNA-loaded RISC is ultimately responsible for sequence-specific gene silencing of target mRNAs236,239. The sense (or “passenger”) strand of the miRNA may be degraded

or used to guide a second RISC. Thus, in some cases, both miRNA duplex strands can

138 direct gene silencing. There are two major mechanisms by which RISC mediates gene silencing: translational inhibition or mRNA transcript degradation. The mechanism of repression is determined by the degree of nucleotide sequence complementarity between the miRNA guide strand and target mRNA, and each pathway is associated with different

RISC components. In general, most miRNAs are only partially complementary to the target mRNAs they control. Indeed, for some miRNA:mRNA interactions, only 7 nt of homology (miRNA nts 2-8; called the “seed region”) is sufficient to cause gene silencing95. This incomplete pairing was initially thought to occur primarily through a translational inhibition mechanism that resulted in reduced protein levels, while target mRNA abundance remained unaffected. However, recent evidence supports that incomplete miRNA base-pairing also causes target mRNA destabilization245. Perfect or near-perfect miRNA:mRNA complementarity across the ~19-25 nt stretch of homology results is typically associated with more robust gene silencing71. The latter is sometimes referred to as an “siRNA” mechanism (see next paragraph for details).

A.3 Artificial Inhibitory RNAs

Artificial inhibitory RNAs can be designed to reduce expression of any gene of interest. There are three major classes, each of which mimics the structure of natural miRNAs at different stages of processing. The first class, small interfering RNAs

(siRNAs), are structurally identical to mature miRNA duplexes71. In various publications, siRNAs may also be referred to as short interfering RNAs, small inhibitory

RNAs, or short inhibitory RNAs. SiRNAs are synthesized in vitro, and upon delivery to cells, require no maturation by miRNA biogenesis machinery, although some longer

139 siRNAs may be processed by Dicer246. The second class of artificial inhibitory RNAs is short hairpin RNAs (shRNAs), which are structurally similar to pre-miRNAs97,247.

Typically, shRNAs are designed with ~21 nt paired sense and antisense sequences connected by an unpaired loop that is often derived from natural miRNA sequences.

ShRNAs are most often expressed intracellularly from DNA expression cassettes and are not cleaved by Drosha. Instead, the transcription start and stop sites are positioned to generate shRNAs with ~2-3 nt 3’ overhangs, thus making them substrates for nuclear export by Exportin-5 and subsequent processing by Dicer and incorporation into RISC.

The third class of artificial inhibitory RNAs, miRNA shuttles (aka artificial miRNAs), mimic pri-miRNAs78,82,244. Like shRNAs, miRNA shuttles are expressed intracellularly from DNA transgenes. MiRNA shuttles typically contain natural miRNA sequences required to direct correct processing, but the natural, mature miRNA duplex in the stem is replaced by artificial sequences derived from a target gene of interest (Fig. A.2).

Following cleavage by Drosha and Dicer, the artificial duplex miRNA resembles an in vitro-synthesized siRNA bearing identical sequences.

Each class of artificial inhibitory RNAs is capable of mediating gene silencing despite entering the pathway at different steps. The main differences between siRNAs and shRNAs or miRNA shuttles are the mode and duration of expression and gene silencing. As noted, siRNAs are chemically synthesized, and must be delivered to target cells using liposomes or other nucleic acid delivery vehicles. Moreover, siRNAs have relatively short half-lives in vivo248. As a result, a single siRNA administration produces only transient gene silencing effects, and long-term target gene suppression requires chronic or repeated siRNA delivery. In contrast, shRNAs and miRNAs are expressed in

140 vivo from DNA-based delivery systems, such as viral vectors or plasmids, which may be

capable of achieving long-term expression and gene silencing after only one

administration249.

A.4 Advantages of miRNA shuttles

ShRNAs were the first generation of expressed artificial inhibitory RNAs applied in vivo. Although several reports demonstrated efficient shRNA-mediated gene silencing98,250,251, recent evidence supports that miRNA shuttles are more predictably

processed, efficient, and safer than shRNAs79,100,105,252. Here, we discuss the advantages

of using miRNA shuttles in RNAi expression vectors, instead of shRNAs.

As previously mentioned, shRNAs and miRNAs are typically expressed in vivo

from DNA expression cassettes but differ in processing events required for maturation.

This differential processing has direct implications for the predictability of the final

product. Drosha cleaves miRNA shuttles but not shRNAs. This is important because the

Drosha cut site directs Dicer binding and cleavage, and together these two events

ultimately determine the sequence of the mature guide strand. Thus, because engineered

miRNA shuttles faithfully recapitulate miRNA structures required for Drosha and Dicer

cleavage, the guide strand is more consistently produced and accurately predicted79,244.

In contrast, the transcription start and termination sites define shRNA 5’ and 3’ ends,

respectively97,247. In a given transcript, these can vary by a few nucleotides, which results

in a mixed population of mature products. Thus, shRNA-derived guide strands may be more variable than similar sequences produced from a miRNA-based delivery system.

141 Because sequence-specificity is a hallmark of RNAi, using a system with the most

accurately predictable final product is desirable.

The differential processing of miRNA shuttles and shRNAs also affects the

methods by which they are expressed. ShRNAs must be positioned near the transcription

start and stop sites to ensure proper maturation and ultimately, gene silencing function.

Such requirements limit the options available for shRNA expression to systems with

well-defined transcription start and stop sequences. Indeed, shRNAs are most typically

expressed from RNA pol III-dependent promoters that have strong ubiquitous and

constitutive expression patterns. In contrast, because Drosha cleavage, not transcription,

defines the critical Exportin-5 and Dicer binding site, miRNA shuttles can be expressed

from any RNA pol II or pol III-dependent promoter, which broadens their usage for

tissue-specific and/or inducible expression applications79,83,244,253-256. Additionally,

miRNA shuttles can be embedded in introns or untranslated regions (UTRs) of coding genes, which allows for co-expression of a reporter gene and a miRNA from the same transcript103,256. By contrast, due to their expression restrictions, shRNAs are not

amenable to such strategies.

Another notable difference between shRNAs and miRNA shuttles is that the

former may be more prone to inducing non-specific toxicity. Although several studies

have shown shRNAs to effectively direct in vivo gene silencing with no overt toxicity,

recent reports have raised safety concerns that are likely related to high-level,

uncontrolled shRNA expression which may cause non-specific gene dysregulation by

disrupting endogenous miRNA biogenesis100,105,251. In one study, vector-delivered

shRNA expression in mouse brain caused robust toxicity in striatal neurons105. However,

142 expressing the identical mature inhibitory RNA sequence in the context of a miRNA

shuttle system driven by the same promoter mitigated this toxicity while still directing

efficient target gene silencing105. This improved safety profile likely resulted from the miRNA shuttles being expressed at lower levels compared to analogous shRNAs.

Although the reasons for these expression differences are unclear, miRNA shuttles more closely mimic natural miRNA structures and may be more efficiently processed and less likely to interfere with endogenous miRNA biogenesis and function.

In summary, miRNA shuttles may be superior to shRNAs because they more closely recapitulate natural miRNA structures, are more predictably processed, are amenable to control by tissue-specific and/or regulated promoters, can be co-expressed

from coding gene introns or UTRs, and may have a better safety profile. For these

reasons, miRNA shuttles are outstanding systems for long-term RNAi studies. In this chapter, we present a rapid method to design and clone miRNA shuttles.

A.5 Design of miRNA shuttles

Our miRNA shuttle cloning system is modified from a human mir-30a (hsa-mir-

30a)-derived strategy originally described by Zeng and Cullen244,257. The natural pri-mir-

30 reference sequence is pictured and the mature antisense mir-30 miRNA produced

following Drosha and Dicer cleavage is 22 nt (Fig. A.2). Our pri-miRNA shuttles are

derived from this sequence with the following modifications: (1) the mature mir-30

duplex sequences are replaced by sequences corresponding to the gene of interest (2) the

pri-miRNA shuttle sequences are slightly shorter than the reference sequence, and (3)

XhoI and SpeI restriction enzymes sites are added for cloning purposes (Fig. A.2; boxed

143 grey). Some natural mir-30 features are maintained, including Drosha and Dicer cut sites

(Fig. A.2; grey and black arrowheads, respectively), and the unpaired loop sequence. In addition, the mismatch located just upstream of the Drosha cut sites should be maintained

for proper processing (Fig. A.2, arrows). Here we describe our method to generate the

miRNA shuttle shown in Figure A.2, which targets the e. coli LacZ transcript. Although our miRNA shuttles are constructed using identical methods based on the current best

“design rules”236,239,258,259, in practice not all miRNAs are functionally equivalent and it is

difficult to predict which sequence will most effectively elicit target gene silencing. We

therefore typically design 4-5 miRNAs per target gene to improve the chance of

generating the most effective miRNA.

For all steps below, refer to Figures A.3 and A.4.

Step 1. Choose 22 nt mature miRNA sequences

There are several publications describing specific rules for optimal inhibitory RNA

design236,239,258,259. Among these, we incorporate three in our miRNA shuttles. First,

RISC preferentially loads the more thermodynamically unstable (i.e. more AU-rich) 5’

end of a miRNA duplex. Thus, to ensure proper loading of the antisense guide strand, the

target site sequence should consist of a GC-rich 5’ end (between positions 3-7) and AU-

rich 3’ end (last 3-4 bases). Positions 1-2 of the target site do not influence the strand-

biasing since they correspond to the 3’ overhang region. Strand-biasing primarily

depends on the “core” duplex region depicted in Figure A.2. The second rule is to select sequences that will yield an antisense guide strand with ~60% or less GC-content, which can be refractory to RNAi-mediated gene silencing. Finally, if using a RNA pol III-based

144 promoter, such as the example described in this method, ensure that there are no long

stretches (4 or more) of tandem T (U in RNA) nucleotides within the mature miRNA

shuttle, since pol III terminates on ~4-5 or greater T’s. Long stretches of T’s will not effect miRNA expression if using a pol II-based promoter, which terminates with a typical poly-adenylation signal (AATAAA). In this example, a 22 nt sequence from the e. coli LacZ gene is embedded in a miRNA. The miRNA guide strand is the reverse complement, and forms perfect Watson-Crick base-pairing with the cognate target.

Step 2. Generate DNA template reference sequences and corresponding miRNA shuttle.

We generate miRNA shuttles using DNA polymerase extension of two annealed DNA

oligonucleotide primers designed to contain all required sequences shown in Figure A.2.

The natural mature miRNA duplex, in this case from hsa-mir-30a, is replaced with the target gene-derived artificial miRNA duplex (Fig. A.2). Regardless of the mature miRNA sequence, each miRNA and therefore, each oligonucleotide, contains common sequence elements so that the cloning method is identical for every miRNA shuttle generated. As a result, primer design is critical for proper miRNA production and cloning. To aid primer design, it is helpful to generate a reference sequence for DNA transcription template used to produce a given miRNA shuttle (Figs A.2 and A.3).

A. The terminal stem (nucleotides 1-12 and 76-86).

Nucleotides 1-12

1. The 5’ end (nts 1-6) contains an XhoI cloning site (5’ CTCGAG 3’), of which

the 3’ GAG nucleotides are derived from natural mir-30a sequences.

145 2. Additional mir-30a-derived stem sequences located at nucleotide positions 7-

12 (5’ TGAGCG 3’), are included in every mir-30a shuttle.

Nucleotides 76-86

1. The T nucleotide at position 76 is immediately 3’ of the Drosha cut site on the

antisense strand

2. Positions 76-83 (5’ TGCCTACT 3’) are mir-30a-derived stem sequences in

which the 3’ terminal ACT nucleotides are the first half of a SpeI restriction

site.

3. The AGT nucleotides located at positions 84-86 are added to generate a 3’

terminal SpeI site (5’ ACTAGT 3’).

B. The artificial miRNA guide strand (nucleotides 54-75)

1. This is the exact guide strand sequence shown in step 1.

2. The guide strand is placed between the Dicer and Drosha cut sites located at

positions 53 and 76 (Fig. A.2; black and grey arrowheads, respectively and

Fig. A.3, Step 2B).

C. The artificial miRNA passenger strand (nucleotides 13-34)

1. The first nucleotide of the passenger strand (position 13) must be mismatched

with position 75 of the guide strand to best mimic mir-30a structure.

a. In our example, note the last guide strand nucleotide at position 75 is

an A.

b. The mismatch as position 13 must then be a C, A, or G, since A:U is a

normal Watson-Crick base pair. The natural has-mir-30a transcript

has an A:C mismatch, but any mismatch will work at this position.

146 c. If the position 75 nucleotide is a C, position 13 must be a U (T in DNA

template), C or A.

d. If the position 75 nucleotide is a U (T in DNA template), position 13

must be a U or C.

e. If the position 75 nucleotide is an A, position 13 must be an A, G or C.

2. The remaining 21 nts of the passenger strand are identical to positions 2-22 of

the mRNA target sequence

D. Loop sequence (nucleotides 35-53)

1. The loop sequence (5’ – CTGTAAAGCCACAGATGGG – 3’) is derived

from hsa-mir-30a with a single modification that facilitates folding in mFOLD

but does not affect function.

2. The two terminal nucleotides (underlined here) at both ends base pair and

form a portion of the miRNA shuttle stem (5’ –

CT36GTGAAGCCACAGATGG52G – 3’), but all other nucleotides remained

unpaired (Figures A1.2 and A1.3). Note again that T36 is a U in the miRNA

transcript, which allows base-pairing with G52.

Steps 3 and 4. Convert the DNA transcription template from step 2 into an RNA molecule (replace T’s with U’s) and then confirm secondary structure using mFOLD.

Among the many free online mFOLD programs available, we have found the www.idtdna.com/Scitools/Applications/mFold/ website to be particularly quick and user friendly. The mFOLD program typically identifies several predicted secondary structures. If correctly designed, the top predicted structure for each miRNA shuttle

147 should have the same features shown in Figure A.2B. The only variance between two different miRNA shuttles will be the duplex miRNA portion. Identify Dicer and Drosha cut sites (Fig. A.2 B) to confirm that the expected guide strand will be excised and that the “core” duplex agrees with strand-biasing rules.

A.6 Primer Design and Cloning of miRNA Shuttles

Step 5: Design primers

After the reference sequences for each miRNA shuttle have been generated, the next step is to design DNA oligonucleotide primers to build the miRNA transcription templates.

Two primers are required per miRNA shuttle (Fig. A.4). As shown, add 4 nucleotides to the 5’ end of each oligonucleotide to provide some additional sequence for efficient XhoI and SpeI restriction enzyme digestion required in Step 8. The forward and reverse primers overlap at the loop sequence, and the unpaired sequences in each primer constitute the miRNA shuttle stem. There are no special oligonucleotide purification requirements; primers can be ordered at low cost from any commercial vendor using the

smallest synthesis scale and simple desalting purity. Reconstitute primers at 1 µg/µl

concentrations in purified water.

Step 6: Generate full-length miRNA shuttles using Taq DNA polymerase.

The annealed loop portion of the forward and reverse primer (see Step 2, D1 above) is

predicted to form an intramolecular hairpin and self-duplex at 37º C, which could impede

the efficiency of the primer extension reaction. However, this natural mir-30 loop

sequence has a predicted melting temperature (Tm) of 56.7º C, which allows us to

148 minimize these potential intramolecular interactions by performing primer extensions at

high temperatures using thermostable Taq polymerase as follows:

5 µl 10X Taq polymerase buffer (Bioline)

1 µl forward primer (1 µg/µl)

1 µl reverse primer (1 µg/µl)

4 µl 2.5 mM dNTPs

1 µl 50 mM MgCl2

0.5 µλ Taq polymerase (5 units/µl)

37.5 µl dH20

Incubate reaction at the following temperatures for one cycle:

94º C for 2 minutes; 52º C for 1 minute; 72º C for 15 minutes.

Step 7: Purify product using a PCR Purification Kit (QIAquick; Qiagen), and elute in 30

µl water.

Step 8: Digest the linear DNA transcription template with XhoI and SpeI:

30 µl eluant

4 µl 10X restriction enzyme buffer (NEB2; New England BioLabs)

1.0 µl SpeI (10 units/µl)

149 1.0 µl XhoI (20 units/µl)

4 µl 10X BSA (10 mg/ml)

Incubate 37º C 4 hours to overnight.

Step 9: Digest U6T6 vector with XhoI and XbaI and gel purify the digested vector and

miRNAs (from Step 9).

Step 10: Ligate miRNA shuttle into U6T6.

To produce a miRNA transcript, the DNA transcription template generated in steps 1-7

must be cloned downstream of a promoter and termination signal. In this example, we

are using the RNA polymerase III-dependent mouse U6 promoter to transcribe miRNA

shuttles. Downstream of the U6 promoter is a pol III termination signal (TTTTTT). The

XhoI – SpeI digested product from step 7 is cloned between the U6 promoter and the

termination signal using XhoI – XbaI restriction sitres. SpeI and XbaI digested ends are compatible. An obvious question is: “why not include XbaI instead of SpeI at the 3’ end of the miRNA template?” Theoretically, and in practice, XbaI – XbaI ligation produces

functional miRNAs. However, we have found that SpeI – XbaI ligation produces more

robust miRNA shuttle expression and gene silencing compared to identical miRNAs

cloned using only XbaI. The reason for this difference is unclear but may be due to slight

secondary structure changes that alter the pri-miRNA transcript stability and/or

processing, therefore affecting the gene silencing efficiency. Thus, the choice of

restriction enzymes and nearby flanking sequences can impact the overall functionality of

the resulting miRNA shuttle. 150

Following digestion, gel purify and quantify the miRNA and U6T6 products, then ligate

and transform into bacteria using standard protocols. U6T6 is kanamycin resistant and

colonies should therefore be screened on LB agar plates containing kanamycin. Because

the U6T6 vector is digested with two non-cohesive enzymes (XhoI and XbaI), no

intramolecular ligation should occur, and we typically do not find kanamycin resistant

colonies on “vector-only” control plates. To further minimize the potential for background, the vector should be treated with alkaline phosphatase after digestion. If little background is evident, we typically pick four colonies per miRNA shuttle construct and grow each of them in liquid culture for DNA miniprep (using QIAquick; Qiagen for transfection quality plasmid).

Step 11A: Identify positive miRNA shuttle clones.

Positive plasmids can be identified by EcoRI restriction enzyme digestion followed by agarose gel electrophoresis (1.5% TBE or TAE gel) and UV imaging of ethidium bromide-stained DNA. Empty U6T6 has three EcoRI sites; one is lost upon XhoI – XbaI digestion (Figure A1.4, Step 8).

Expected EcoRI products for successfully cloned U6T6 miRNA expression cassettes:

Empty U6T6 U6T6 with miRNA shuttle

3437 bp (contains vector) 3437 bp (contains vector)

151 383 bp (contains U6 promoter) 489 bp (contains entire U6 miRNA cassette)

83 bp (contains T6 terminator)

Step 11B: Sequence U6.miRNA plasmids containing the correct EcoRI digestion pattern.

Sequence two positive clones from each miRNA expression cassette to ensure that errors in primer synthesis or extension did not occur. We use ABI dye terminator sequencing and an ABI 3130xl sequencer (Applied Biosystems). Sequencing reactions of some miRNA hairpin plasmids prematurely terminate because of strong secondary structures.

To prevent this, include 4% dimethyl sulfoxide (DMSO) in the sequencing reaction.

A.7 Rapid Screen to Identify Functional MiRNA Shuttles

As mentioned, although the “design rules” are good guidelines for miRNA construction, it is difficult to predict which sequence will most effectively direct target gene silencing, if at all. Functional validation is therefore required. Typically, miRNA shuttle plasmids are transfected into a cell line expressing the target gene of interest naturally or from an exogenously delivered expression plasmid. If an easily transfected cell line expressing your gene of interest is available, perhaps the easiest method to determine gene silencing efficacy is to measure target gene expression using standard biochemical or molecular techniques like western blotting or real-time PCR, one or two days post-transfection. If easily transfected cell lines expressing your gene of interest are not available, another option is to use the dual luciferase assay system we describe here.

This method circumvents transfection-related inconsistencies, which can affect

152 identification of effective miRNA shuttles. This system, modified from a commercial

vector, provides rapid, consistent, and quantitative measurement of gene silencing, regardless of transfection efficiency, due mostly to an internal transfection control. Our method requires: (1) generation of a luciferase reporter plasmid in which the target gene of interest is cloned as the 3’ untranslated region (UTR) of Renilla luciferase; (2) co-

transfection of miRNA shuttles with this reporter into any available cell line; and (3) a

dual luciferase assay kit (Promega). In the following paragraph, we describe the general

principles underlying this method.

First, we created a new dual luciferase vector called pSICHECK-DEST, which we

modified from the commercial pSICHECK2 vector (Promega). Our new construct

contains Gateway® destination vector sequences (DEST®; Invitrogen) that allow for

rapid recombination of Gateway ENTR® (Invitrogen) vectors containing target genes of

interest (Fig. A.5). These gene-specific reporters are generated rapidly using LR Clonase

enzyme (Invitrogen). The end result is a reporter plasmid (pSICHECK-GOI) expressing a fusion transcript in which the target gene of interest is inserted downstream of the

Renilla luciferase stop codon, effectively making it the Renilla luciferase 3’ untranslated

region (3’ UTR; Fig. A.5). From this fusion transcript, only the Renilla luciferase coding

region is translated into protein (Fig. A.6). An artificial miRNA that silences the target

gene of interest will also reduce Renilla luciferase protein expression (Fig. A.6).

Importantly, pSICHECK-GOI also contains a separate HSV-TK promoter driven firefly

luciferase gene, which is unaffected by the artificial miRNA and therefore serves as a

transfection control (Fig. A.6). Gene silencing can then determined by measuring Renilla

153 and firefly luciferase activity in lysates from transfected cells using a dual luciferase assay kit (Promega).

Generating the Luciferase Reporter Plasmid

The pSICHECK-DEST vector allows rapid recombination of Gateway ENTR vectors containing target gene sequences into the 3’ untranslated (3’ UTR) portion of the Renilla luciferase gene. Importantly, many Gateway®-ready cDNAs are commercially available

(e.g. Open Biosystems, Origene, Addgene, PlasmID). If an ENTR vector containing your gene of interest is not available through a commercial resource, one can be generated using standard PCR techniques and TOPO® cloning kits (pENTR/D-TOPO; Invitrogen).

To create a pSICHECK vector containing your gene of interest, mix 150 ng pSICHECK-

DEST with 150 ng pENTR-GOI and perform an LR reaction, using manufacturer’s instructions (LR Clonase; Invitrogen). There are several layers of selection to ensure efficient recombination. First, a successful LR reaction will replace the ccdB and chloramphenicol resistance genes in pSICHECK-DEST with the gene of interest. The bacterial ccdB gene is toxic to typical E. coli used in competent cells (e.g. TOP10,

DH5α) and, if expressed, bacterial colonies cannot grow. Additionally, pSICHECK-

DEST also expresses the ampicillin (Amp) resistance gene, while typical ENTR vectors are kanamycin resistant, thereby preventing growth of any unrecombined ENTR vector colonies. Any living, ampicillin resistant bacterial colonies arising from LR recombination are likely correct, but restriction digests using, for example BamHI (Fig.

A.5), should be performed on miniprep DNA to confirm correctness of the plasmids.

154 Dual luciferase assay to determine gene silencing efficiency of artificial miRNA shuttles

For in vitro miRNA screening assays, we use HEK293s because of their rapid growth and ease of transfection using liposome reagents (e.g. Lipofectamine-2000; Invitrogen).

Moreover, HEK293s possess all miRNA biogenesis and gene silencing machinery and have been used extensively for RNAi studies. Seed 2.5 X 104 HEK293 cells one day prior to transfection on 96 well plates. Since the goal of the experiment is to identify the most potent miRNA shuttle at the lowest dose, we typically perform screens using two different ratios of miRNA to target vector. Controls can include empty U6T6 and/or a non-targeting miRNA shuttle. If your gene of interest is not E. coli LacZ, the example miLacZ described in this chapter could be used as a non-targeting control. Table 1 shows plasmid amounts used in a typical transfection for a miRNA efficacy screen. Perform each transfection in triplicate wells. 48 hours later, measure Renilla and firefly luciferase activities using a dual luciferase reporter assay kit (Promega) following manufacturer’s instructions. Determine gene silencing by plotting the ratio of Renilla to firefly luciferase activity in triplicate samples and normalize to “empty U6T6” or “control miRNA” data.

This method provides a quick and consistent assay to identify lead miRNA shuttle vectors targeting your gene of interest, which is its major advantage. Results from this reporter assay are consistently translatable to natural transcripts, but we still advise that gene silencing be confirmed using a second method, such as western blotting or real-time

PCR.

A.8 Summary

155 RNA interference is a powerful tool for investigating basic biological questions

and developing nucleic acid therapeutics. Artificial miRNA shuttles can be engineered to

suppress any gene of interest using designs that closely mimic natural miRNA structures.

As such, miRNA shuttles are excellent systems for achieving long-term gene silencing in many different cell and animal models. For plasmid or vector-based RNAi expression systems, miRNA shuttles are arguably the best available system.

156 Table A.1. Example transfection worksheet for miRNA efficacy screen using pSICHECK-GOI

Target Plasmid miRNA plasmid pSICHECK-GOI (ng) miLacZ control (ng) miGOI (ng) Empty U6T6 (ng) 5 0 0 195 5 25 0 170 5 195 0 0 5 0 25 170 5 0 195 0 0 0 0 200

157 Figure A.1. MiRNA biogenesis pathway. Natural miRNAs are first transcribed from

genomic DNA as primary miRNA trancripts (pri-miRNAs), which form hairpin

structures. In the nucleus, the pri-miRNA is cleaved at one end by the Microprocessor

complex, which contains Drosha and DGCR8 proteins. This cut produces a shorter pre- miRNA hairpin containing a 2 nucleotide 3’ overhang. Exportin-5 (EXP5) then shuttles the pre-miRNA to the cytoplasm through the nuclear pore. There, a second cut, mediated by the Dicer enzyme, removes the loop region to produce a mature duplex miRNA containing di-nucleotide 3’ overhangs at both ends. One strand (the antisense “guide” strand) incorporates into a riboprotein multimer called the RNA Induced Silencing

Complex (RISC), which then directs sequence-specific gene silencing of target

messenger RNAs. Artificial miRNAs are designed to enter the pathway as pri-miRNAs in the nucleus, and therefore undergo all processing steps required to produce a mature miRNA.

158

Figure A.1

159

Figure A.2. Artificial miRNA shuttles derived from natural human mir-30a. (a)

Sequence and structure of human mir-30a. (b) An artificial miRNA shuttle, based on human mir-30a, targeting the E. coli LacZ gene. Grey and black arrowheads indicate

Drosha and Dicer cut sites, respectively. The mature guide strand is underlined. In B,

XhoI and SpeI cloning sites are boxed. Further details are provided in the text.

160

Figure A.3. Designing and cloning miRNA shuttle vectors: Steps 1-4. All details are described in the text.

161

Figure A.4. Designing and cloning miRNA shuttle vectors: Steps 5-11. All details are described in the text.

162 Figure A.5. Generating a luciferase reporter plasmid with the Gateway® system. pSICHECK-DEST is a new dual luciferase vector we created to facilitate cloning of luciferase-gene of interest fusion transcripts. Details are described in the text.

Abbreviations: SV40, SV40 promoter; Luc, luciferase; Chloramp, chloramphenicol resistance gene; HSV-TK, herpes simplex virus thymidine kinase promoter; Kan, kanamycin resistance gene; Amp, ampicillin resistance gene; GOI, gene of interest.

163

Figure A.5 164

Figure A.6. Overview of luciferase assay screen for testing miRNA gene silencing.

The pSICHECK-GOI plasmid, once constructed as described, expresses two different luciferase reporter genes. The first is a Firefly luciferase reporter used to normalize transfections, and the second is a hybrid transcript in which the Renilla luciferase mRNA is fused with the target gene of interest (GOI). A STOP codon is placed between the

Renilla coding region and the GOI sequence, so that a fusion mRNA is produced, but not a fusion protein. The GOI sequences are thus analogous to the 3’ UTR region of Renilla luciferase. Any inhibitory RNA (e.g. one arising from the U6 miRNA plasmid depicted here) that effectively silences the target GOI will therefore Renilla luciferase expression, which can then be quantified using commercially available kits. Details are described in the text.

165

APPENDIX B

GFP, A WORD OF WARNING

B.1 Introduction

Green Fluorescent Protein (GFP), originally discovered in the jellyfish Aequorea

victoria, has become one of the most important tools in modern biology. Indeed, the

2008 Nobel Prize in Chemistry was awarded to Shimomura, Chalfie, and Tsien for their

discovery and development of GFP as a biological reporter gene. Over the years, numerous variants of the wild-type GFP (wtGFP) protein were created to improve stability and brightness, and optimize expression in mammalian cells. The most commonly used variant is enhanced GFP (eGFP), which was codon-optimized for mammalian cell expression (humanized), and engineered with a serine-65 to threonine mutation that made it 35 times brighter than the wtGFP protein260. This eGFP variant has

proven important in numerous areas, including the gene therapy field, where it has been a

critical tool for optimizing vector-mediated delivery approaches.

The utility of GFP as a biological reporter, and some data suggesting it could be cytotoxic, spurred the optimization of alternative fluorescent proteins from other

organisms, including the sea pansy, Renilla reniformis. A humanized form of Renilla

GFP (hrGFP) was introduced to market several years ago and billed as a less toxic

alternative to Aequorea-derived GFP proteins 261. Our lab is interested in expressing non- 166 coding RNAs to muscle using adeno-associated viral (AAV) vectors, which cannot be directly visualized in living animals. Therefore, as an indirect indicator of AAV transduction and non-coding RNA expression, we incorporated eGFP or hrGFP reporter cassettes into our AAV backbones. The goal of this study was to histologically assess the potential toxicity of AAV6-delivered eGFP and hrGFP in adult mouse muscle. We found that mouse muscles tolerated high levels of eGFP markedly better than comparable levels of hrGFP across a range of dosages. These data indicate that eGFP is less toxic than hrGFP in adult mouse muscle.

B.2 Materials and Methods

Adeno-associated virus (AAV) production. The AAV.CMV.hrGFP and

AAV.CMV.eGFP vectors were cloned as previously described(ITRs)98,262, and contained a CMV promoter and SV40 polyadenylation signal flanked by two AAV2 inverted terminal repeats. AAV serotype 6 (AAV6) vectors were generated by the Viral Vector

Core Facility at TRINCH. Titers were determined by quantitative PCR and reported as

DNAse Resistant Particles (DRP). AAV6.CMV.hrGFP, 4.8 X 1012 total DRP and

AAV6.CMV.eGFP, 5.2 X 1012 total DRP. Endotoxin was reported in endotoxin units

(EU)/ml. AAV6.CMV.hrGFP 0.507 EU/ml and AAV6.CMV.eGFP 0.493 EU/ml.

Mice. All animal studies were approved by Institutional Animal Care and Use

Committee at the Research Institute at Nationwide Children’s Hospital (TRINCH).

Female C57BL/6 mice were purchased from the National Cancer Institute, and Jackson

Laboratory.

167 AAV injections. Six to eight-week-old C57BL/6 females received a 50 microliter intramuscular (IM) injection into the TA containing AAV.CMV.hrGFP or

AAV.CMV.eGFP at one of multiple doses: 8 X 108, 3 X 109, 8 X 109, 3 X 1010, or 1 X

1011DRP units. In vivo transduction was determined using a fluorescent dissecting

microscope (M165FC, Leica).

Histological analysis. TA muscles were dissected from IM injected mice at 1, 2, 3, or 4 weeks post-injection for histological analysis (n=4 muscles per group at each time point for each dose). 10-µm cryosections were generated and H&E stained as previously described 166.

B.3 Results and discussion

Our lab is interested in expressing therapeutic non-coding RNAs to muscle using

AAV vectors. Indeed we have demonstrated potential for a novel muscle gene therapy

for dominant muscular dystrophies207. In these studies we originally incorporated hrGFP

reporter cassettes into our AAV backbones because tracking transduction is necessary,

and hrGFP was suggested to be less toxic than eGFP. The reporter gene serves as an

indirect indicator of our therapeutic microRNA (miRNA) expression. Though reports of

toxicity from high levels of GFP existed169,263, we never observed this phenomenon when

delivering hrGFP to neonatal mouse muscle207. However, when using hrGFP as an

injection control in adult muscle we saw signs of toxicity including inflammatory

infiltrates and centrally located nuclei when delivered at high doses. To find the safest

optimal dose of AAV.CMV.hrGFP to deliver to adult muscle we performed a dose

escalation study and injected mice with a range of titers including: 8 X 108, 3 X 109, 8 X

168 109 and 3 X 1010DRP units. By analyzing the muscle histology weekly (up to 4 weeks),

we found a direct correlation between dose and time in relation to toxicity (Fig. B.1).

Concentrating on the 4 week time point (Fig. B.1 a) we identified hallmarks of myotoxicity that increased with dose escalation. Specifically, AAV.CMV.hrGFP titers of

8 X 108 and 3 X 109DRP showed no sign of toxicity at any time-point tested. However, when the dose was increased to 8 X 109DRP, the muscles began to show signs of immune cell infiltration and myofiber degeneration (Fig. B.1 a). In comparison, virtually all myofibers from muscles injected with 3 X 1010DRP contained abundant central nuclei

indicating there had been degeneration and subsequent regeneration (Fig. B.1 a). Earlier

time points at this dose also indicate time as a toxicity factor (Fig. B.1 b). Similar to the

low range titers in the dose escalation study, at 1-week post injection 3 X 1010DRP of

AAV.CMV.hrGFP results in histologically normal muscle. However, as time increases so does the histopathology. The 2-week time point is marked by moderate degeneration,

the 3-week by a combination of degeneration and regeneration (Fig. B.1 b). As stated

earlier, the 4-week time point is consistently marked by complete regeneration (Fig. B.1

b). These were surprising results considering hrGFP is marketed as less toxic then eGFP.

While this may be a cell specific scenario, various other muscle gene therapy publications

have utilized eGFP vectors at significantly higher doses for longer time points264. To

ensure the results seen in the hrGFP dose escalation were not a side effect of our injection

or viral preparation methods, we injected both AAV.CMV.hrGFP and AAV.CMV.eGFP

unilaterally into the same mouse at increasing doses up to 1 X 1011DRP. Our results

confirmed that mouse muscles tolerated high levels of eGFP markedly better than comparable levels of hrGFP (Fig. B.2 a). However, at the 4-week time point even the

169 eGFP muscles began to show mild signs of toxicity at the highest dose tested (Fig. B.2 a).

It is possible these histopathological features also existed in previous studies using this dose but were concealed by the already dystrophic muscle of the animal models.

Importantly, both GFP species showed equally bright expression in the transduced muscles at all time points (Fig. B.2 b), even through the processes of degeneration and regeneration (Fig. B.2 c). These results combined with the fact that the

AAV.CMV.hrGFP and AAV.CMV.eGFP viral preparations contained virtually equivalent levels of acceptable endotoxin (Table B.1), confirm that eGFP is less toxic than hrGFP when delivered to adult mouse muscle. Like our previous studies, many potential gene therapies are screened with the aid of GFP as a reporter gene. The results presented here serve as a word of warning, because dose escalation and long term time points are important data for many preclinical therapeutic studies. In this light, it is extremely important to choose the safest reporter gene to ensure the elimination of potential false negative results. For adult muscle gene therapy, this study supports the use of eGFP.

170 Table B.1. Endotoxin report

171 Figure B.1 HrGFP toxicity is dependent on dose and time. (a) H&E staining shows at a 4 week time point escalating doses of hrGFP induce myotoxicity. Low hrGFP vector doses (8 X108 and 3 X 109) were histologically normal but higher hrGFP vector doses (8

X 109 and 3 X 1010 DRP) were capable of massive myofiber degeneration, mononuclear

cell infiltration and complete muscle turnover as indicated by myofibers with central

nuclei. (b) hrGFP-induced toxicity is influenced by time. Representative sections from a

3 X 1010 DRP dose are shown over a 4-week period. 1 week post-injection muscles are

histologically normal but toxicity increases substantially at 2, 3 and 4 weeks post

injection. Scale bars, 50 µm.

172

Figure B.1 173 Figure B.2. EGFP is better tolerated than hrGFP in muscle. (a) Representative fields

depict the highest degree of histopathology observed in muscles injected with 1 x

1011DRP of AAV.CMV.hrGFP or eGFP at specified times. (b) hrGFP and eGFP epifluorescence shows AAV6 transduction of adult TA 1-week and 4-weeks post

injection. (c) 4-weeks post injection a cross section of hrGFP transduced myofibers

confirming abundant expression even in the regenerative state of the muscle. Scale bars,

50 µm.

174

Figure B.2

175

References

1. INSERM, F.M.o.H.a. Prevalence of Rare Diseases: Bibliographic Data. in Orphanet Report, Vol. 2008 (Orphanet, 2008). 2. Padberg, G.W. Thesis, Leiden University (1982). 3. Tyler, F.H. & Stephens, F.E. Studies in disorders of muscle. II Clinical manifestations and inheritance of facioscapulohumeral dystrophy in a large family. Ann Intern Med 32, 640-660 (1950). 4. Padberg, G.W., Lunt, P.W., Koch, M. & Fardeau, M. Diagnostic criteria for facioscapulohumeral muscular dystrophy. Neuromuscul Disord 1, 231-234 (1991). 5. Trevisan, C.P., et al. Facioscapulohumeral muscular dystrophy and occurrence of heart arrhythmia. Eur Neurol 56, 1-5 (2006). 6. Faustmann, P.M., et al. Cardiac involvement in facio-scapulo-humeral muscular dystrophy: a family study using Thallium-201 single-photon-emission-computed tomography. J Neurol Sci 144, 59-63 (1996). 7. Galetta, F., et al. Subclinical cardiac involvement in patients with facioscapulohumeral muscular dystrophy. Neuromuscul Disord 15, 403-408 (2005). 8. Tsuji, M., Kinoshita, M., Imai, Y., Kawamoto, M. & Kohara, N. Facioscapulohumeral muscular dystrophy presenting with hypertrophic cardiomyopathy: a case study. Neuromuscul Disord 19, 140-142 (2009). 9. Pandya, S., King, W.M. & Tawil, R. Facioscapulohumeral dystrophy. Phys Ther 88, 105-113 (2008). 10. Tawil, R. & Van Der Maarel, S.M. Facioscapulohumeral muscular dystrophy. Muscle Nerve 34, 1-15 (2006). 11. van der Maarel, S.M., Frants, R.R. & Padberg, G.W. Facioscapulohumeral muscular dystrophy. Biochim Biophys Acta 1772, 186-194 (2007). 12. Wohlgemuth, M., et al. Dysphagia in facioscapulohumeral muscular dystrophy. Neurology 66, 1926-1928 (2006). 13. Olsen, D.B., Gideon, P., Jeppesen, T.D. & Vissing, J. Leg muscle involvement in facioscapulohumeral muscular dystrophy assessed by MRI. J Neurol 253, 1437- 1441 (2006). 14. A prospective, quantitative study of the natural history of facioscapulohumeral muscular dystrophy (FSHD): implications for therapeutic trials. The FSH-DY Group. Neurology 48, 38-46 (1997).

176 15. D'Angelo, M.G., et al. Respiratory pattern in an adult population of dystrophic patients. J Neurol Sci 306, 54-61 (2011). 16. Padberg, G.W., et al. On the significance of retinal vascular disease and hearing loss in facioscapulohumeral muscular dystrophy. Muscle Nerve 2, S73-80 (1995). 17. Saito, Y., et al. Facioscapulohumeral muscular dystrophy with severe mental retardation and epilepsy. Brain Dev 29, 231-233 (2007). 18. Jewesbury, R.C. Facio-scapulo-humeral Muscular Dystrophy (Landouzy- Dejerine) Type. Proc R Soc Med 23, 1114-1115 (1930). 19. Wijmenga, C., et al. Location of facioscapulohumeral muscular dystrophy gene on chromosome 4. Lancet 336, 651-653 (1990). 20. Wijmenga, C., et al. Chromosome 4q DNA rearrangements associated with facioscapulohumeral muscular dystrophy. Nat Genet 2, 26-30 (1992). 21. Lyle, R., Wright, T.J., Clark, L.N. & Hewitt, J.E. The FSHD-associated repeat, D4Z4, is a member of a dispersed family of homeobox-containing repeats, subsets of which are clustered on the short arms of the acrocentric chromosomes. Genomics 28, 389-397 (1995). 22. Zhang, Y., Forner, J., Fournet, S. & Jeanpierre, M. Improved characterization of FSHD mutations. Ann Genet 44, 105-110 (2001). 23. Lemmers, R.J.L., et al. Complete allele information in the diagnosis of facioscapulohumeral muscular dystrophy by triple DNA analysis. Ann Neurol 50, 816-819 (2001). 24. Kohler, J., Rohrig, D., Bathke, K.D. & Koch, M.C. Evaluation of the facioscapulohumeral muscular dystrophy (FSHD1) phenotype in correlation to the concurrence of 4q35 and 10q26 fragments. Clin Genet 55, 88-94 (1999). 25. Tupler, R., et al. Monosomy of distal 4q does not cause facioscapulohumeral muscular dystrophy. J Med Genet 33, 366-370 (1996). 26. Ding, H., et al. Characterization of a double homeodomain protein (DUX1) encoded by a cDNA homologous to 3.3 kb dispersed repeated elements. Hum Mol Genet 7, 1681-1694 (1998). 27. Gabriels, J., et al. Nucleotide sequence of the partially deleted D4Z4 locus in a patient with FSHD identifies a putative gene within each 3.3 kb element. Gene 236, 25-32 (1999). 28. Hewitt, J.E., et al. Analysis of the tandem repeat locus D4Z4 associated with facioscapulohumeral muscular dystrophy. Human molecular genetics 3, 1287- 1295 (1994). 29. Lunt, P.W. 44th ENMC International Workshop: Facioscapulohumeral Muscular Dystrophy: Molecular Studies 19-21 July 1996, Naarden, The Netherlands. Neuromuscul Disord 8, 126-130 (1998). 30. van Geel, M., et al. Genomic analysis of human chromosome 10q and 4q telomeres suggests a common origin. Genomics 79, 210-217 (2002). 31. Lemmers, R.J., et al. Worldwide population analysis of the 4q and 10q subtelomeres identifies only four discrete interchromosomal sequence transfers in human evolution. Am J Hum Genet 86, 364-377 (2010). 32. Lemmers, R.J., et al. A Unifying Genetic Model for Facioscapulohumeral Muscular Dystrophy. Science (2010).

177 33. Spurlock, G., Jim, H.P. & Upadhyaya, M. Confirmation that the specific SSLP microsatellite allele 4qA161 segregates with fascioscapulohumeral muscular dystrophy (FSHD) in a cohort of multiplex and simplex FSHD families. Muscle Nerve 42, 820-821 (2010). 34. Bodega, B., et al. Remodeling of the chromatin structure of the facioscapulohumeral muscular dystrophy (FSHD) locus and upregulation of FSHD-related gene 1 (FRG1) expression during human myogenic differentiation. BMC Biol 7, 41 (2009). 35. Gabellini, D., Green, M.R. & Tupler, R. Inappropriate gene activation in FSHD: a repressor complex binds a chromosomal repeat deleted in dystrophic muscle. Cell 110, 339-348 (2002). 36. van Overveld, P.G., et al. Hypomethylation of D4Z4 in 4q-linked and non-4q- linked facioscapulohumeral muscular dystrophy. Nat Genet 35, 315-317 (2003). 37. Zeng, W., et al. Specific loss of histone H3 lysine 9 trimethylation and HP1gamma/cohesin binding at D4Z4 repeats is associated with facioscapulohumeral dystrophy (FSHD). PLoS Genet 5, e1000559 (2009). 38. van Deutekom, J.C., et al. Identification of the first gene (FRG1) from the FSHD region on human chromosome 4q35. Hum Mol Genet 5, 581-590 (1996). 39. Gabellini, D., et al. Facioscapulohumeral muscular dystrophy in mice overexpressing FRG1. Nature 439, 973-977 (2006). 40. Hanel, M.L., Wuebbles, R.D. & Jones, P.L. Muscular dystrophy candidate gene FRG1 is critical for muscle development. Dev Dyn 238, 1502-1512 (2009). 41. Liu, Q., Jones, T.I., Tang, V.W., Brieher, W.M. & Jones, P.L. Facioscapulohumeral muscular dystrophy region gene-1 (FRG-1) is an actin- bundling protein associated with muscle-attachment sites. J Cell Sci 123, 1116- 1123 (2010). 42. Wuebbles, R.D., Hanel, M.L. & Jones, P.L. FSHD region gene 1 (FRG1) is crucial for angiogenesis linking FRG1 to facioscapulohumeral muscular dystrophy-associated vasculopathy. Dis Model Mech 2, 267-274 (2009). 43. Chen, S.C., et al. Decreased proliferation kinetics of mouse myoblasts overexpressing FRG1. PLoS One 6, e19780 (2011). 44. Sun, C.Y., et al. Facioscapulohumeral muscular dystrophy region gene 1 is a dynamic RNA-associated and actin-bundling protein. J Mol Biol 411, 397-416 (2011). 45. van Koningsbruggen, S., et al. FRG1P-mediated aggregation of proteins involved in pre-mRNA processing. Chromosoma 116, 53-64 (2007). 46. Laoudj-Chenivesse, D., et al. Increased levels of adenine nucleotide translocator 1 protein and response to oxidative stress are early events in facioscapulohumeral muscular dystrophy muscle. J Mol Med 83, 216-224 (2005). 47. Arashiro, P., et al. Transcriptional regulation differs in affected facioscapulohumeral muscular dystrophy patients compared to asymptomatic related carriers. Proc Natl Acad Sci U S A 106, 6220-6225 (2009). 48. Masny, P.S., et al. Analysis of allele-specific RNA transcription in FSHD by RNA-DNA FISH in single myonuclei. Eur J Hum Genet 18, 448-456 (2010).

178 49. Osborne, R.J., Welle, S., Venance, S.L., Thornton, C.A. & Tawil, R. Expression profile of FSHD supports a link between retinal vasculopathy and muscular dystrophy. Neurology 68, 569-577 (2007). 50. Winokur, S.T., et al. Expression profiling of FSHD muscle supports a defect in specific stages of myogenic differentiation. Hum Mol Genet 12, 2895-2907 (2003). 51. Snider, L., et al. Facioscapulohumeral dystrophy: incomplete suppression of a retrotransposed gene. PLoS Genet 6, e1001181 (2010). 52. Dixit, M., et al. DUX4, a candidate gene of facioscapulohumeral muscular dystrophy, encodes a transcriptional activator of PITX1. Proc Natl Acad Sci U S A 104, 18157-18162 (2007). 53. Kowaljow, V., et al. The DUX4 gene at the FSHD1A locus encodes a pro- apoptotic protein. Neuromuscul Disord 17, 611-623 (2007). 54. Clapp, J., et al. Evolutionary conservation of a coding function for D4Z4, the tandem DNA repeat mutated in facioscapulohumeral muscular dystrophy. Am J Hum Genet 81, 264-279 (2007). 55. Bosnakovski, D., et al. An isogenetic myoblast expression screen identifies DUX4-mediated FSHD-associated molecular pathologies. Embo J 27, 2766-2779 (2008). 56. Wallace, L.M., et al. DUX4, a candidate gene for facioscapulohumeral muscular dystrophy, causes p53-dependent myopathy in vivo. Ann Neurol (2010). 57. Kissel, J.T., et al. Pilot trial of albuterol in facioscapulohumeral muscular dystrophy. FSH-DY Group. Neurology 50, 1402-1406 (1998). 58. Tawil, R., et al. A pilot trial of prednisone in facioscapulohumeral muscular dystrophy. FSH-DY Group. Neurology 48, 46-49 (1997). 59. Wagner, K.R., et al. A phase I/IItrial of MYO-029 in adult subjects with muscular dystrophy. Ann Neurol 63, 561-571 (2008). 60. Kissel, J.T., et al. Randomized, double-blind, placebo-controlled trial of albuterol in facioscapulohumeral dystrophy. Neurology 57, 1434-1440 (2001). 61. http://clinicaltrials.gov/ct2/results?term=FSHD. 62. McPherron, A.C., Lawler, A.M. & Lee, S.J. Regulation of skeletal muscle mass in mice by a new TGF-beta superfamily member. Nature 387, 83-90 (1997). 63. Lee, S.J. & McPherron, A.C. Regulation of myostatin activity and muscle growth. Proc Natl Acad Sci U S A 98, 9306-9311 (2001). 64. Barnett, N.D., et al. Winging of the scapula: the underlying biomechanics and an orthotic solution. Proc Inst Mech Eng H 209, 215-223 (1995). 65. Mummery, C.J., Copeland, S.A. & Rose, M.R. Scapular fixation in muscular dystrophy. Cochrane Database Syst Rev, CD003278 (2003). 66. NIH. National Institutes of Health Office of Rare Diseases Research. in Rare Diseases and Related Terms (http://rarediseases.info.nih.gov/RareDiseaseList.aspx?PageID=1) (2009). 67. Harper, P.S. Myotonic Dystrophy (2nd ed.), (W.B. Saunders, London, 1989). 68. Flanigan, K.M., et al. Genetic characterization of a large, historically significant Utah kindred with facioscapulohumeral dystrophy. Neuromuscul Disord 11, 525- 529 (2001).

179 69. Fire, A., et al. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391, 806-811 (1998). 70. Bartel, D.P. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 116, 281-297 (2004). 71. Elbashir, S.M., et al. Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 411, 494-498 (2001). 72. Elbashir, S.M., Lendeckel, W. & Tuschl, T. RNA interference is mediated by 21- and 22-nucleotide RNAs. Genes Dev 15, 188-200 (2001). 73. Lai, E.C. Micro RNAs are complementary to 3' UTR sequence motifs that mediate negative post-transcriptional regulation. Nat Genet 30, 363-364 (2002). 74. Lagos-Quintana, M., Rauhut, R., Lendeckel, W. & Tuschl, T. Identification of novel genes coding for small expressed RNAs. Science 294, 853-858 (2001). 75. Lagos-Quintana, M., Rauhut, R., Meyer, J., Borkhardt, A. & Tuschl, T. New microRNAs from mouse and human. RNA 9, 175-179 (2003). 76. Molnar, A., et al. Highly specific gene silencing by artificial microRNAs in the unicellular alga Chlamydomonas reinhardtii. Plant J (2009). 77. Molnar, A., Schwach, F., Studholme, D.J., Thuenemann, E.C. & Baulcombe, D.C. miRNAs control gene expression in the single-cell alga Chlamydomonas reinhardtii. Nature 447, 1126-1129 (2007). 78. Zeng, Y., Cai, X. & Cullen, B.R. Use of RNA polymerase II to transcribe artificial microRNAs. Methods Enzymol 392, 371-380 (2005). 79. Boudreau, R.L., Martins, I. & Davidson, B.L. Artificial microRNAs as siRNA shuttles: improved safety as compared to shRNAs in vitro and in vivo. Mol Ther 17, 169-175 (2009). 80. Lagos-Quintana, M., et al. Identification of tissue-specific microRNAs from mouse. Curr Biol 12, 735-739 (2002). 81. Borchert, G.M., Lanier, W. & Davidson, B.L. RNA polymerase III transcribes human microRNAs. Nat Struct Mol Biol 13, 1097-1101 (2006). 82. Cai, X., Hagedorn, C.H. & Cullen, B.R. Human microRNAs are processed from capped, polyadenylated transcripts that can also function as mRNAs. RNA 10, 1957-1966 (2004). 83. Lee, Y., et al. MicroRNA genes are transcribed by RNA polymerase II. Embo J 23, 4051-4060 (2004). 84. Gregory, R.I., et al. The Microprocessor complex mediates the genesis of microRNAs. Nature 432, 235-240 (2004). 85. Han, J., et al. The Drosha-DGCR8 complex in primary microRNA processing. Genes Dev 18, 3016-3027 (2004). 86. Landthaler, M., Yalcin, A. & Tuschl, T. The human DiGeorge syndrome critical region gene 8 and Its D. melanogaster homolog are required for miRNA biogenesis. Curr Biol 14, 2162-2167 (2004). 87. Lee, Y., et al. The nuclear RNase III Drosha initiates microRNA processing. Nature 425, 415-419 (2003). 88. Han, J., et al. Molecular basis for the recognition of primary microRNAs by the Drosha-DGCR8 complex. Cell 125, 887-901 (2006).

180 89. Zeng, Y. & Cullen, B.R. Structural requirements for pre-microRNA binding and nuclear export by Exportin 5. Nucleic Acids Res 32, 4776-4785 (2004). 90. Provost, P., et al. Ribonuclease activity and RNA binding of recombinant human Dicer. EMBO J 21, 5864-5874 (2002). 91. Zhang, H., Kolb, F.A., Brondani, V., Billy, E. & Filipowicz, W. Human Dicer preferentially cleaves dsRNAs at their termini without a requirement for ATP. EMBO J 21, 5875-5885 (2002). 92. Zhang, H., Kolb, F.A., Jaskiewicz, L., Westhof, E. & Filipowicz, W. Single processing center models for human Dicer and bacterial RNase III. Cell 118, 57- 68 (2004). 93. Matranga, C., Tomari, Y., Shin, C., Bartel, D.P. & Zamore, P.D. Passenger-strand cleavage facilitates assembly of siRNA into Ago2-containing RNAi enzyme complexes. Cell 123, 607-620 (2005). 94. Ro, S., Park, C., Young, D., Sanders, K.M. & Yan, W. Tissue-dependent paired expression of miRNAs. Nucleic Acids Res 35, 5944-5953 (2007). 95. Lewis, B.P., Burge, C.B. & Bartel, D.P. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 120, 15-20 (2005). 96. Rose, S.D., et al. Functional polarity is introduced by Dicer processing of short substrate RNAs. Nucleic Acids Res 33, 4140-4156 (2005). 97. Paddison, P.J., Caudy, A.A., Bernstein, E., Hannon, G.J. & Conklin, D.S. Short hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian cells. Genes Dev 16, 948-958 (2002). 98. Harper, S.Q., et al. RNA interference improves motor and neuropathological abnormalities in a Huntington's disease mouse model. Proc Natl Acad Sci U S A 102, 5820-5825 (2005). 99. Xia, H., et al. RNAi suppresses polyglutamine-induced neurodegeneration in a model of spinocerebellar ataxia. Nat Med 10, 816-820 (2004). 100. Grimm, D., et al. Fatality in mice due to oversaturation of cellular microRNA/short hairpin RNA pathways. Nature 441, 537-541 (2006). 101. Fechner, H., et al. Cardiac-targeted RNA interference mediated by an AAV9 vector improves cardiac function in coxsackievirus B3 cardiomyopathy. J Mol Med 86, 987-997 (2008). 102. Du, G., Yonekubo, J., Zeng, Y., Osisami, M. & Frohman, M.A. Design of expression vectors for RNA interference based on miRNAs and RNA splicing. FEBS J 273, 5421-5427 (2006). 103. Harper, S.Q., et al. Optimization of feline immunodeficiency virus vectors for RNA interference. J Virol 80, 9371-9380 (2006). 104. Li, M.J., et al. Inhibition of HIV-1 infection by lentiviral vectors expressing Pol III-promoted anti-HIV RNAs. Mol Ther 8, 196-206 (2003). 105. McBride, J.L., et al. Artificial miRNAs mitigate shRNA-mediated toxicity in the brain: implications for the therapeutic development of RNAi. Proc Natl Acad Sci U S A 105, 5868-5873 (2008).

181 106. Mook, O.R., Baas, F., de Wissel, M.B. & Fluiter, K. Allele-specific cancer cell killing in vitro and in vivo targeting a single-nucleotide polymorphism in POLR2A. Cancer Gene Ther (2009). 107. Saydam, O., et al. Herpes simplex virus 1 amplicon vector-mediated siRNA targeting epidermal growth factor receptor inhibits growth of human glioma cells in vivo. Mol Ther 12, 803-812 (2005). 108. Quinzii, C.M., et al. X-linked dominant scapuloperoneal myopathy is due to a mutation in the gene encoding four-and-a-half-LIM protein 1. Am J Hum Genet 82, 208-213 (2008). 109. Foroud, T., et al. A mutation in myotilin causes spheroid body myopathy. Neurology 65, 1936-1940 (2005). 110. Garvey, S.M., Miller, S.E., Claflin, D.R., Faulkner, J.A. & Hauser, M.A. Transgenic mice expressing the myotilin T57I mutation unite the pathology associated with LGMD1A and MFM. Hum Mol Genet 15, 2348-2362 (2006). 111. Hauser, M.A., et al. myotilin Mutation found in second pedigree with LGMD1A. Am J Hum Genet 71, 1428-1432 (2002). 112. Hauser, M.A., et al. Myotilin is mutated in limb girdle muscular dystrophy 1A. Hum Mol Genet 9, 2141-2147 (2000). 113. Schroder, R., et al. Beyond LGMD1A: myotilin is a component of central core lesions and nemaline rods. Neuromuscul Disord 13, 451-455 (2003). 114. Selcen, D. & Engel, A.G. Mutations in myotilin cause myofibrillar myopathy. Neurology 62, 1363-1371 (2004). 115. Moza, M., et al. Targeted deletion of the muscular dystrophy gene myotilin does not perturb muscle structure or function in mice. Mol Cell Biol 27, 244-252 (2007). 116. Corbett, M.A., et al. An alphaTropomyosin mutation alters dimer preference in nemaline myopathy. Ann Neurol 57, 42-49 (2005). 117. de Haan, A., van der Vliet, M.R., Gommans, I.M., Hardeman, E.C. & van Engelen, B.G. Skeletal muscle of mice with a mutation in slow alpha-tropomyosin is weaker at lower lengths. Neuromuscul Disord 12, 952-957 (2002). 118. Durling, H.J., et al. De novo missense mutation in a constitutively expressed exon of the slow alpha-tropomyosin gene TPM3 associated with an atypical, sporadic case of nemaline myopathy. Neuromuscul Disord 12, 947-951 (2002). 119. Ilkovski, B., et al. Disease severity and thin filament regulation in M9R TPM3 nemaline myopathy. J Neuropathol Exp Neurol 67, 867-877 (2008). 120. Kee, A.J. & Hardeman, E.C. Tropomyosins in skeletal muscle diseases. Adv Exp Med Biol 644, 143-157 (2008). 121. Laing, N.G., et al. A mutation in the alpha tropomyosin gene TPM3 associated with autosomal dominant nemaline myopathy NEM1. Nat Genet 10, 249 (1995). 122. Penisson-Besnier, I., Monnier, N., Toutain, A., Dubas, F. & Laing, N. A second pedigree with autosomal dominant nemaline myopathy caused by TPM3 mutation: a clinical and pathological study. Neuromuscul Disord 17, 330-337 (2007).

182 123. Tan, P., et al. Homozygosity for a nonsense mutation in the alpha-tropomyosin slow gene TPM3 in a patient with severe infantile nemaline myopathy. Neuromuscul Disord 9, 573-579 (1999). 124. Lehtokari, V.L., et al. Identification of a founder mutation in TPM3 in nemaline myopathy patients of Turkish origin. Eur J Hum Genet 16, 1055-1061 (2008). 125. Rethinasamy, P., et al. Molecular and physiological effects of alpha-tropomyosin ablation in the mouse. Circ Res 82, 116-123 (1998). 126. Carbone, I., et al. Mutation in the CAV3 gene causes partial caveolin-3 deficiency and hyperCKemia. Neurology 54, 1373-1376 (2000). 127. Galbiati, F., et al. Caveolin-3 null mice show a loss of caveolae, changes in the microdomain distribution of the dystrophin-glycoprotein complex, and t-tubule abnormalities. J Biol Chem 276, 21425-21433 (2001). 128. Minetti, C., et al. Mutations in the caveolin-3 gene cause autosomal dominant limb-girdle muscular dystrophy. Nat Genet 18, 365-368 (1998). 129. Traverso, M., et al. Caveolin-3 T78M and T78K missense mutations lead to different phenotypes in vivo and in vitro. Lab Invest 88, 275-283 (2008). 130. Tang, G. siRNA and miRNA: an insight into RISCs. Trends Biochem Sci 30, 106- 114 (2005). 131. Kurosawa, T., Igarashi, S., Nishizawa, M. & Onodera, O. Selective silencing of a mutant transthyretin allele by small interfering RNAs. Biochem Biophys Res Commun 337, 1012-1018 (2005). 132. Miller, V.M., Paulson, H.L. & Gonzalez-Alegre, P. RNA interference in neuroscience: progress and challenges. Cell Mol Neurobiol 25, 1195-1207 (2005). 133. Miller, V.M., et al. Allele-specific silencing of dominant disease genes. Proc Natl Acad Sci U S A 100, 7195-7200 (2003). 134. Rodriguez-Lebron, E. & Paulson, H.L. Allele-specific RNA interference for neurological disease. Gene Ther 13, 576-581 (2006). 135. Schwarz, D.S., et al. Designing siRNA that distinguish between genes that differ by a single nucleotide. PLoS Genet 2, e140 (2006). 136. Cho, D.H. & Tapscott, S.J. Myotonic dystrophy: emerging mechanisms for DM1 and DM2. Biochim Biophys Acta 1772, 195-204 (2007). 137. Lee, H.C., et al. Abnormal Na channel gating in murine cardiac myocytes deficient in myotonic dystrophy protein kinase. Physiol Genomics 12, 147-157 (2003). 138. Mounsey, J.P., Mistry, D.J., Ai, C.W., Reddy, S. & Moorman, J.R. Skeletal muscle sodium channel gating in mice deficient in myotonic dystrophy protein kinase. Hum Mol Genet 9, 2313-2320 (2000). 139. Berul, C.I., et al. DMPK dosage alterations result in atrioventricular conduction abnormalities in a mouse myotonic dystrophy model. J Clin Invest 103, R1-7 (1999). 140. Reddy, S., et al. Mice lacking the myotonic dystrophy protein kinase develop a late onset progressive myopathy. Nat Genet 13, 325-335 (1996). 141. Lee, Y., Jeon, K., Lee, J.T., Kim, S. & Kim, V.N. MicroRNA maturation: stepwise processing and subcellular localization. EMBO J 21, 4663-4670 (2002).

183 142. Langlois, M.A., et al. Cytoplasmic and nuclear retained DMPK mRNAs are targets for RNA interference in myotonic dystrophy cells. J Biol Chem 280, 16949-16954 (2005). 143. Ohrt, T., et al. Fluorescence correlation spectroscopy and fluorescence cross- correlation spectroscopy reveal the cytoplasmic origination of loaded nuclear RISC in vivo in human cells. Nucleic Acids Res 36, 6439-6449 (2008). 144. Robb, G.B., Brown, K.M., Khurana, J. & Rana, T.M. Specific and potent RNAi in the nucleus of human cells. Nat Struct Mol Biol 12, 133-137 (2005). 145. Weinmann, L., et al. Importin 8 is a gene silencing factor that targets argonaute proteins to distinct mRNAs. Cell 136, 496-507 (2009). 146. Celegato, B., et al. Parallel protein and transcript profiles of FSHD patient muscles correlate to the D4Z4 arrangement and reveal a common impairment of slow to fast fibre differentiation and a general deregulation of MyoD-dependent genes. Proteomics 6, 5303-5321 (2006). 147. Eisenberg, I., et al. Distinctive patterns of microRNA expression in primary muscular disorders. Proc Natl Acad Sci U S A 104, 17016-17021 (2007). 148. Reed, P.W., Corse, A.M., Porter, N.C., Flanigan, K.M. & Bloch, R.J. Abnormal expression of mu-crystallin in facioscapulohumeral muscular dystrophy. Exp Neurol 205, 583-586 (2007). 149. van Deutekom, J.C., et al. FSHD associated DNA rearrangements are due to deletions of integral copies of a 3.2 kb tandemly repeated unit. Hum Mol Genet 2, 2037-2042 (1993). 150. Flanigan, K.M. Facioscapulohumeral Muscular Dystrophy and Scapuloperoneal Disorders. in Myology, Third Edition, Vol. 2 1123-1133 (McGraw-Hill Medical Publishing Division, New York, 2004). 151. Klinge, L., et al. Severe phenotype in infantile facioscapulohumeral muscular dystrophy. Neuromuscul Disord 16, 553-558 (2006). 152. Bakker, E., et al. The FSHD-linked locus D4F104S1 (p13E-11) on 4q35 has a homologue on 10qter. Muscle Nerve 2, S39-44 (1995). 153. Lemmers, R.J., et al. Inter- and intrachromosomal sub-telomeric rearrangements on 4q35: implications for facioscapulohumeral muscular dystrophy (FSHD) aetiology and diagnosis. Hum Mol Genet 7, 1207-1214 (1998). 154. Lemmers, R.J., et al. Specific sequence variations within the 4q35 region are associated with facioscapulohumeral muscular dystrophy. Am J Hum Genet 81, 884-894 (2007). 155. Thomas, N.S., et al. A large patient study confirming that facioscapulohumeral muscular dystrophy (FSHD) disease expression is almost exclusively associated with an FSHD locus located on a 4qA-defined 4qter subtelomere. J Med Genet 44, 215-218 (2007). 156. de Greef, J.C., Frants, R.R. & van der Maarel, S.M. Epigenetic mechanisms of facioscapulohumeral muscular dystrophy. Mutat Res 647, 94-102 (2008). 157. de Greef, J.C., et al. Common epigenetic changes of D4Z4 in contraction- dependent and contraction-independent FSHD. Hum Mutat (2009). 158. Hewitt, J.E., et al. Analysis of the tandem repeat locus D4Z4 associated with facioscapulohumeral muscular dystrophy. Hum Mol Genet 3, 1287-1295 (1994).

184 159. Jiang, G., et al. Testing the position-effect variegation hypothesis for facioscapulohumeral muscular dystrophy by analysis of histone modification and gene expression in subtelomeric 4q. Hum Mol Genet 12, 2909-2921 (2003). 160. Snider, L., et al. RNA transcripts, miRNA-sized fragments and proteins produced from D4Z4 units: new candidates for the pathophysiology of facioscapulohumeral dystrophy. Hum Mol Genet 18, 2414-2430 (2009). 161. Wuebbles, R.D., Long, S.W., Hanel, M.L. & Jones, P.L. Testing the effects of FSHD candidate gene expression in vertebrate muscle development. Int J Clin Exp Pathol 3, 386-400 (2010). 162. Sandri, M., et al. Caspase 3 expression correlates with skeletal muscle apoptosis in Duchenne and facioscapulo human muscular dystrophy. A potential target for pharmacological treatment? J Neuropathol Exp Neurol 60, 302-312 (2001). 163. Ostlund, C., Garcia-Carrasquillo, R.M., Belayew, A. & Worman, H.J. Intracellular trafficking and dynamics of double homeodomain proteins. Biochemistry 44, 2378-2384 (2005). 164. Kawakami, K., et al. A transposon-mediated gene trap approach identifies developmentally regulated genes in zebrafish. Dev Cell 7, 133-144 (2004). 165. Jacks, T., et al. Tumor spectrum analysis in p53-mutant mice. Curr Biol 4, 1-7 (1994). 166. Harper, S.Q., et al. Modular flexibility of dystrophin: implications for gene therapy of Duchenne muscular dystrophy. Nat Med 8, 253-261 (2002). 167. Thornhill, P., Bassett, D., Lochmuller, H., Bushby, K. & Straub, V. Developmental defects in a zebrafish model for muscular dystrophies associated with the loss of fukutin-related protein (FKRP). Brain 131, 1551-1561 (2008). 168. Salva, M.Z., et al. Design of tissue-specific regulatory cassettes for high-level rAAV-mediated expression in skeletal and cardiac muscle. Mol Ther 15, 320-329 (2007). 169. Huang, W.Y., Aramburu, J., Douglas, P.S. & Izumo, S. Transgenic expression of green fluorescent protein can cause dilated cardiomyopathy. Nat Med 6, 482-483 (2000). 170. Franich, N.R., et al. AAV vector-mediated RNAi of mutant huntingtin expression is neuroprotective in a novel genetic rat model of Huntington's disease. Mol Ther 16, 947-956 (2008). 171. Wang, F., et al. AAV-mediated expression of vascular endothelial growth factor induces choroidal neovascularization in rat. Invest Ophthalmol Vis Sci 44, 781- 790 (2003). 172. Blankinship, M.J., et al. Efficient transduction of skeletal muscle using vectors based on adeno-associated virus serotype 6. Mol Ther 10, 671-678 (2004). 173. Wang, Z., et al. Adeno-associated virus serotype 8 efficiently delivers genes to muscle and heart. Nat Biotechnol 23, 321-328 (2005). 174. Chipuk, J.E., et al. Direct activation of Bax by p53 mediates mitochondrial membrane permeabilization and apoptosis. Science 303, 1010-1014 (2004). 175. Degli Esposti, M. & Dive, C. Mitochondrial membrane permeabilisation by Bax/Bak. Biochem Biophys Res Commun 304, 455-461 (2003).

185 176. Gupta, S., Radha, V., Furukawa, Y. & Swarup, G. Direct transcriptional activation of human caspase-1 by tumor suppressor p53. J Biol Chem 276, 10585- 10588 (2001). 177. Sadasivam, S., et al. Caspase-1 activator Ipaf is a p53-inducible gene involved in apoptosis. Oncogene 24, 627-636 (2005). 178. Thalappilly, S., Sadasivam, S., Radha, V. & Swarup, G. Involvement of caspase 1 and its activator Ipaf upstream of mitochondrial events in apoptosis. FEBS J 273, 2766-2778 (2006). 179. de Greef, J.C., et al. Hypomethylation is restricted to the D4Z4 repeat array in phenotypic FSHD. Neurology 69, 1018-1026 (2007). 180. Rijkers, T., et al. FRG2, an FSHD candidate gene, is transcriptionally upregulated in differentiating primary myoblast cultures of FSHD patients. J Med Genet 41, 826-836 (2004). 181. Winokur, S.T., et al. Facioscapulohumeral muscular dystrophy (FSHD) myoblasts demonstrate increased susceptibility to oxidative stress. Neuromuscul Disord 13, 322-333 (2003). 182. Xia, S.J., Pressey, J.G. & Barr, F.G. Molecular pathogenesis of rhabdomyosarcoma. Cancer Biol Ther 1, 97-104 (2002). 183. Heng, M.Y., Detloff, P.J. & Albin, R.L. Rodent genetic models of Huntington disease. Neurobiol Dis 32, 1-9 (2008). 184. Liu, D.X. & Lobie, P.E. Transcriptional activation of p53 by Pitx1. Cell Death Differ 14, 1893-1907 (2007). 185. Fernando, P., Kelly, J.F., Balazsi, K., Slack, R.S. & Megeney, L.A. Caspase 3 activity is required for skeletal muscle differentiation. Proc Natl Acad Sci U S A 99, 11025-11030 (2002). 186. Larsen, B.D., et al. Caspase 3/caspase-activated DNase promote cell differentiation by inducing DNA strand breaks. Proc Natl Acad Sci U S A 107, 4230-4235 (2010). 187. Koenig, M., et al. Complete cloning of the Duchenne muscular dystrophy (DMD) cDNA and preliminary genomic organization of the DMD gene in normal and affected individuals. Cell 50, 509-517 (1987). 188. Scott, J.M., et al. Viral vectors for gene transfer of micro-, mini-, or full-length dystrophin. Neuromuscul Disord 12 Suppl 1, S23-29 (2002). 189. Lim, L.E. & Campbell, K.P. The sarcoglycan complex in limb-girdle muscular dystrophy. Curr Opin Neurol 11, 443-452 (1998). 190. Li, J., et al. rAAV vector-mediated sarcogylcan gene transfer in a hamster model for limb girdle muscular dystrophy. Gene Ther 6, 74-82 (1999). 191. Cordier, L., et al. Rescue of skeletal muscles of gamma-sarcoglycan-deficient mice with adeno-associated virus-mediated gene transfer. Mol Ther 1, 119-129 (2000). 192. Goyenvalle, A., et al. Rescue of dystrophic muscle through U7 snRNA-mediated exon skipping. Science 306, 1796-1799 (2004). 193. Gregorevic, P., et al. Systemic delivery of genes to striated muscles using adeno- associated viral vectors. Nat Med 10, 828-834 (2004).

186 194. Mah, C., et al. Physiological correction of Pompe disease by systemic delivery of adeno-associated virus serotype 1 vectors. Mol Ther 15, 501-507 (2007). 195. Pacak, C.A., et al. Long-term skeletal muscle protection after gene transfer in a mouse model of LGMD-2D. Mol Ther 15, 1775-1781 (2007). 196. Yue, Y., et al. A single intravenous injection of adeno-associated virus serotype-9 leads to whole body skeletal muscle transduction in dogs. Mol Ther 16, 1944- 1952 (2008). 197. Kota, J., et al. Follistatin gene delivery enhances muscle growth and strength in nonhuman primates. Sci Transl Med 1, 6ra15 (2009). 198. Rodino-Klapac, L.R., Montgomery, C.L., Mendell, J.R. & Chicoine, L.G. AAV- mediated gene therapy to the isolated limb in rhesus macaques. Methods Mol Biol 709, 287-298 (2011). 199. Mendell, J.R., et al. Sustained alpha-sarcoglycan gene expression after gene transfer in limb-girdle muscular dystrophy, type 2D. Ann Neurol 68, 629-638 (2010). 200. Wallace, L.M., Garwick S.E., and Harper S.Q. RNAi Therapy for Dominant Muscular Dystrophies and Other Myopathies. in Muscle Gene Therapy (ed. Duan, D.) 99-115 (Springer, New York, 2010). 201. Boudreau, R.L. & Davidson, B.L. RNAi therapeutics for CNS disorders. Brain Res 1338, 112-121 (2010). 202. Hanel, M.L., et al. Facioscapulohumeral muscular dystrophy (FSHD) region gene 1 (FRG1) is a dynamic nuclear and sarcomeric protein. Differentiation 81, 107- 118 (2011). 203. Tawil, R. Facioscapulohumeral muscular dystrophy. Neurotherapeutics 5, 601- 606 (2008). 204. Bosnakovski, D., Daughters, R.S., Xu, Z., Slack, J.M. & Kyba, M. Biphasic myopathic phenotype of mouse DUX, an ORF within conserved FSHD-related repeats. PLoS One 4, e7003 (2009). 205. Padberg, G., et al. Linkage studies in autosomal dominant facioscapulohumeral muscular dystrophy. Journal of the neurological sciences 65, 261-268 (1984). 206. Boudreau, R.L., Garwick-Coppens, S.E., Liu, J., Wallace, L.M., and Harper S.Q. Rapid cloning and validation of microRNA shuttle vectors: A practical guide. in RNA Interference Techniques, Vol. 1 (ed. Harper, S.Q.) 19-37 (Humana Press, New York, 2011). 207. Wallace, L.M., Garwick-Coppens, S. E., Tupler, R., and Harper S. Q. RNA interference improves myopathic phenotypes in mice over-expressing FSHD region gene 1 (FRG1). Mol Ther (2011). 208. Bortolanza, S., et al. AAV6-mediated Systemic shRNA Delivery Reverses Disease in a Mouse Model of Facioscapulohumeral Muscular Dystrophy. Molecular therapy : the journal of the American Society of Gene Therapy 19, 2055-2064 (2011). 209. Kunkel, L.M., et al. Analysis of deletions in DNA from patients with Becker and Duchenne muscular dystrophy. Nature 322, 73-77 (1986). 210. Sicinski, P., et al. The molecular basis of muscular dystrophy in the mdx mouse: a point mutation. Science 244, 1578-1580 (1989).

187 211. Ryder-Cook, A.S., et al. Localization of the mdx mutation within the mouse dystrophin gene. The EMBO journal 7, 3017-3021 (1988). 212. Goyenvalle, A., Seto, J.T., Davies, K.E. & Chamberlain, J. Therapeutic approaches to muscular dystrophy. Human molecular genetics 20, R69-78 (2011). 213. Vanderplanck, C., et al. The FSHD Atrophic Myotube Phenotype Is Caused by DUX4 Expression. PloS one 6, e26820 (2011). 214. Mendell, J.R., et al. Limb-girdle muscular dystrophy type 2D gene therapy restores alpha-sarcoglycan and associated proteins. Ann Neurol 66, 290-297 (2009). 215. Chen, Y.H., Chang, M. & Davidson, B.L. Molecular signatures of disease brain endothelia provide new sites for CNS-directed enzyme therapy. Nature medicine 15, 1215-1218 (2009). 216. Koerber, J.T., Maheshri, N., Kaspar, B.K. & Schaffer, D.V. Construction of diverse adeno-associated viral libraries for directed evolution of enhanced gene delivery vehicles. Nature protocols 1, 701-706 (2006). 217. Richards, M., Coppee, F., Thomas, N., Belayew, A. & Upadhyaya, M. Facioscapulohumeral muscular dystrophy (FSHD): an enigma unravelled? Hum Genet 131, 325-340 (2012). 218. Lefebvre, S., et al. Identification and characterization of a spinal muscular atrophy-determining gene. Cell 80, 155-165 (1995). 219. Melki, J., et al. De novo and inherited deletions of the 5q13 region in spinal muscular atrophies. Science 264, 1474-1477 (1994). 220. Lefebvre, S., et al. Correlation between severity and SMN protein level in spinal muscular atrophy. Nat Genet 16, 265-269 (1997). 221. McAndrew, P.E., et al. Identification of proximal spinal muscular atrophy carriers and patients by analysis of SMNT and SMNC gene copy number. Am J Hum Genet 60, 1411-1422 (1997). 222. Hua, Y. & Krainer, A.R. Antisense-mediated exon inclusion. Methods Mol Biol 867, 307-323 (2012). 223. Sleigh, J.N., Gillingwater, T.H. & Talbot, K. The contribution of mouse models to understanding the pathogenesis of spinal muscular atrophy. Dis Model Mech 4, 457-467 (2011). 224. Wadman, R.I., et al. Drug treatment for spinal muscular atrophy types II and III. Cochrane Database Syst Rev, CD006282 (2011). 225. Wadman, R.I., et al. Drug treatment for spinal muscular atrophy type I. Cochrane Database Syst Rev, CD006281 (2011). 226. Dominguez, E., et al. Intravenous scAAV9 delivery of a codon-optimized SMN1 sequence rescues SMA mice. Hum Mol Genet 20, 681-693 (2011). 227. Foust, K.D., et al. Rescue of the spinal muscular atrophy phenotype in a mouse model by early postnatal delivery of SMN. Nat Biotechnol 28, 271-274 (2010). 228. Le, T.T., et al. Temporal requirement for high SMN expression in SMA mice. Hum Mol Genet 20, 3578-3591 (2011). 229. Porensky, P.N., et al. A single administration of morpholino antisense oligomer rescues spinal muscular atrophy in mouse. Hum Mol Genet 21, 1625-1638 (2012).

188 230. Valori, C.F., et al. Systemic delivery of scAAV9 expressing SMN prolongs survival in a model of spinal muscular atrophy. Sci Transl Med 2, 35ra42 (2010). 231. Fire, A.Z. Gene silencing by double-stranded RNA. Cell Death Differ 14, 1998- 2012 (2007). 232. Lee, R.C. & Ambros, V. An extensive class of small RNAs in Caenorhabditis elegans. Science 294, 862-864 (2001). 233. Odling-Smee, L. Complex set of RNAs found in simple green algae. Nature 447, 518 (2007). 234. Hutvagner, G., et al. A cellular function for the RNA-interference enzyme Dicer in the maturation of the let-7 small temporal RNA. Science 293, 834-838 (2001). 235. Hutvagner, G. & Zamore, P.D. A microRNA in a multiple-turnover RNAi enzyme complex. Science 297, 2056-2060 (2002). 236. Khvorova, A., Reynolds, A. & Jayasena, S.D. Functional siRNAs and miRNAs exhibit strand bias. Cell 115, 209-216 (2003). 237. Macrae, I.J., et al. Structural basis for double-stranded RNA processing by Dicer. Science 311, 195-198 (2006). 238. Marques, J.T., et al. A structural basis for discriminating between self and nonself double-stranded RNAs in mammalian cells. Nat Biotechnol 24, 559-565 (2006). 239. Schwarz, D.S., et al. Asymmetry in the assembly of the RNAi enzyme complex. Cell 115, 199-208 (2003). 240. Tolia, N.H. & Joshua-Tor, L. Slicer and the argonautes. Nat Chem Biol 3, 36-43 (2007). 241. Wang, Y., Medvid, R., Melton, C., Jaenisch, R. & Blelloch, R. DGCR8 is essential for microRNA biogenesis and silencing of embryonic stem cell self- renewal. Nat Genet 39, 380-385 (2007). 242. Yi, R., Qin, Y., Macara, I.G. & Cullen, B.R. Exportin-5 mediates the nuclear export of pre-microRNAs and short hairpin RNAs. Genes Dev 17, 3011-3016 (2003). 243. Davidson, B.L. & Paulson, H.L. Molecular medicine for the brain: silencing of disease genes with RNA interference. Lancet Neurol 3, 145-149 (2004). 244. Zeng, Y., Wagner, E.J. & Cullen, B.R. Both natural and designed micro RNAs can inhibit the expression of cognate mRNAs when expressed in human cells. Mol Cell 9, 1327-1333 (2002). 245. Guo, H., Ingolia, N.T., Weissman, J.S. & Bartel, D.P. Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature 466, 835-840 (2010). 246. Amarzguioui, M., et al. Rational design and in vitro and in vivo delivery of Dicer substrate siRNA. Nat Protoc 1, 508-517 (2006). 247. Harper, S.Q. & Davidson, B.L. Plasmid-based RNA interference: construction of small-hairpin RNA expression vectors. Methods Mol Biol 309, 219-235 (2005). 248. Judge, A.D., Bola, G., Lee, A.C. & MacLachlan, I. Design of noninflammatory synthetic siRNA mediating potent gene silencing in vivo. Mol Ther 13, 494-505 (2006). 249. Davidson, B.L. & Harper, S.Q. Viral delivery of recombinant short hairpin RNAs. Methods Enzymol 392, 145-173 (2005).

189 250. McCaffrey, A.P., et al. Inhibition of hepatitis B virus in mice by RNA interference. Nat Biotechnol 21, 639-644 (2003). 251. Rodriguez-Lebron, E., Denovan-Wright, E.M., Nash, K., Lewin, A.S. & Mandel, R.J. Intrastriatal rAAV-mediated delivery of anti-huntingtin shRNAs induces partial reversal of disease progression in R6/1 Huntington's disease transgenic mice. Mol Ther 12, 618-633 (2005). 252. Boudreau, R.L., et al. Nonallele-specific silencing of mutant and wild-type huntingtin demonstrates therapeutic efficacy in Huntington's disease mice. Mol Ther 17, 1053-1063 (2009). 253. Silva, J.M., et al. Second-generation shRNA libraries covering the mouse and human genomes. Nat Genet 37, 1281-1288 (2005). 254. Stegmeier, F., Hu, G., Rickles, R.J., Hannon, G.J. & Elledge, S.J. A lentiviral microRNA-based system for single-copy polymerase II-regulated RNA interference in mammalian cells. Proc Natl Acad Sci U S A 102, 13212-13217 (2005). 255. Stern, P., et al. A system for Cre-regulated RNA interference in vivo. Proc Natl Acad Sci U S A 105, 13895-13900 (2008). 256. Xia, X.G., Zhou, H., Samper, E., Melov, S. & Xu, Z. Pol II-expressed shRNA knocks down Sod2 gene expression and causes phenotypes of the gene knockout in mice. PLoS Genet 2, e10 (2006). 257. Zeng, Y. & Cullen, B.R. Sequence requirements for micro RNA processing and function in human cells. RNA 9, 112-123 (2003). 258. Reynolds, A., et al. Rational siRNA design for RNA interference. Nat Biotechnol 22, 326-330 (2004). 259. Li, L., Lin, X., Khvorova, A., Fesik, S.W. & Shen, Y. Defining the optimal parameters for hairpin-based knockdown constructs. RNA 13, 1765-1774 (2007). 260. Zhang, G., Gurtu, V. & Kain, S.R. An enhanced green fluorescent protein allows sensitive detection of gene transfer in mammalian cells. Biochem Biophys Res Commun 227, 707-711 (1996). 261. www.chem-agilent.com/pdf/strata/240031.pdf. 262. Allaire, P.D., et al. Connecdenn, a novel DENN domain-containing protein of neuronal clathrin-coated vesicles functioning in synaptic vesicle endocytosis. J Neurosci 26, 13202-13212 (2006). 263. Taghizadeh, R.R. & Sherley, J.L. CFP and YFP, but not GFP, provide stable fluorescent marking of rat hepatic adult stem cells. J Biomed Biotechnol 2008, 453590 (2008). 264. Rodino-Klapac, L.R., et al. A translational approach for limb vascular delivery of the micro-dystrophin gene without high volume or high pressure for treatment of Duchenne muscular dystrophy. J Transl Med 5, 45 (2007).

190