Identification and Characterization of Novel Genetic Factors Involved in the Differentation and Regeneration of Inner Ear Cells

Research Collection

Doctoral Thesis

Identification and characterization of novel genetic factors involved in the differentation and regeneration of inner ear cells

Author(s): Riccardi, Sabrina

Publication Date: 2015

Permanent Link: https://doi.org/10.3929/ethz-a-010439511

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library

DISS. ETH NO. 22545

IDENTIFICATION AND CHARACTERIZATION OF NOVEL GENETIC FACTORS INVOLVED IN THE DIFFERENTIATION AND REGENERATION OF INNER EAR CELLS

A dissertation submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

SABRINA RICCARDI

Master of Science ETH in Medicinal and Industrial Pharmaceutical Sciences, ETH Zürich

Born on 26.03.1983

citizen of

Pratteln (BL), Switzerland

Accepted on the recommendation of

Prof. Dr. Michael Detmar

Prof. Dr. Jonathan Hall

Dr. Bernd Kinzel

2015

Table of Contents

1. Summary ...... 4 1.1 Summary ...... 4 1.2 Zusammenfassung ...... 6 2. Introduction ...... 9 2.1 Hearing loss: Prevalence, causes and medical need ...... 9 2.2 Anatomy of the inner ear and the organ of Corti ...... 11 2.3 Molecular mechanisms of inner ear hair cells development and regeneration ...... 13 2.3.1 Sensory hair cell development ...... 13 2.3.2 Mechanisms of hair cell regeneration ...... 16 2.5 The importance of microRNAs in inner ear hair cell development ...... 19 2.6 In vitro and ex‐vivo models for assessment of hair cell differentiation ...... 21 2.6.1 The inner ear cell line UB/OC‐1 ...... 21 2.6.2 Ex‐vivo assay with organ of Corti explants ...... 24 3 Aim and outline of this dissertion ...... 26 4 Identification and characterization of novel genetic factors involved in the regeneration of inner ear hair cells ...... 28 4.1 Mir‐210 promotes ectopic sensory hair cell formation via transdifferentiation of supporting cells in the organ of Corti ...... 28 4.1.1 Abstract ...... 28 4.1.2 Introduction ...... 29 4.1.3 Material and Methods ...... 33 4.1.4 Results ...... 41 4.1.5 Discussion ...... 55 5. Identification and characterization of Atoh1 modulators ...... 60 5.1 Mir‐152 can epigenetically modulate and reactivate the transcription factor Atoh1 ...... 60

5.1.1 Abstract ...... 60 5.1.2 Introduction ...... 61 5.1.3 Materials and methods ...... 64 5.1.4 Results ...... 74 5.1.5 Discussion ...... 81 6 CONCLUSIONS AND OUTLOOK ...... 84 7 ACKNOWLEDGMENTS ...... 90 8 REFERENCES ...... 92 9 CURRICULUM VITAE ...... 97

1. Summary

1.1 Summary

Hearing loss is the most common sensory defect in developed countries, with more than 5% of people in industrialized nations having significant hearing problems. In most cases, the cause of hearing loss is related to the deterioration and death of hair cells and their associated spiral ganglion neurons. However, despite this, few studies have reported regeneration of the auditory system. Significant gaps remain in our knowledge regarding the molecular pathways underpinning auditory function, including the factors required for cellular regeneration and regulation of cochlear gene expression. Therefore, although there is a massive social and economic demand to develop therapeutic treatments for hearing loss, deafness remains one of the most widespread, costly and poorly understood disabilities in the world.

The aim of the present work was to identify novel genetic factors playing a role in the differentiation and regeneration of inner ear hair cells. This will support the discovery of relevant targets and molecular pathways that might open up new possibilities for the treatment of hearing loss.

To identify genetic elements involved in the differentiation of the sensory hair cells, differential miRNA expression during differentiation of the inner ear progenitor cell line UB/OC‐1 was analyzed. Functional characterization of several miRNAs identified by comprehensive small RNA next generation sequencing revealed one candidate, miR‐210, whose knock‐down triggered differentiation from a progenitor cell stage towards a more differentiated hair cell phenotype.

Since inhibition of miR‐210 in UB/OC‐1 cells changed cell fate from proliferation to differentiation it was reasoned that miR‐210 plays an active role in maintaining the proliferative

4 progenitor cell stage. To evaluate this hypothesis, explants of the mouse organ of Corti was transduced with an adenovirus expressing miR‐210. Ectopic expression of miR‐210 resulted in new hair cell formation in the organ of Corti explants. Analysis of cell fate in a cell lineage tracing model revealed that the formation of new hair cells occurred by transdifferentiation from former Sox2 positive supporting epithelial cells. To better understand the mechanisms of new hair cell formation induced by miR‐210 expression, a target prediction analysis using

TargetScan was performed and compared with previously published miR‐210 interactions retrieved from Metabase. The analysis revealed 18 novel candidate targets besides targets predicted previously. To identify candidate genes facilitating miR‐210 mediated transdifferentiation, immunoprecipitation of RISC complexes enriched for miR‐210 targets was performed, followed by quantitative PCR analysis. Besides the identification of several transcripts that are not yet linked to transdifferentiation or sensory epithelium differentiation, several miR‐210 regulated transcripts that are known to play a role in this mechanism were identified. These results suggest that miR‐210 might be a potential new factor for hearing loss therapy. Moreover, identification of pathways regulated by miR‐210 in the inner ear could reveal novel drug targets for the treatment of hearing loss.

In the second part of the present work, a genetic screen was conducted to identify Atoh1 modulators. Atoh1 is a transcription factor known to be involved in the development and regeneration of sensory hair cells in the organ of Corti. Using a human intestinal epithelial cell line (SW480), engineered with a luciferase reporter gene under the control of the Atoh1 regulatory locus, a small library of 32 miRNA expressing plasmids was screened. This led to the identification of miR‐152 acting as a factor leading to reactivation of Atoh1 gene expression.

Reactivation and expression of Atoh1 was associated with reduced DNA methylation of the enhancer region in the Atoh1 locus. These results are consistent with the suggestion that miR‐

152 acts by inhibiting expression of DNA‐methyltransferases. The link between miR‐152 and

Atoh1 activation in the inner ear organ of Corti was further characterized. MiR‐152 was overexpressed in cochlear explants using a viral vector. No activation in the Atoh1 supporting cells was seen in the organ of Corti explants after overexpression of miR‐152. These results suggest that miR‐152 activates Atoh1 via epigenetic modulation in the intestinal colon cancer cell line but shows no effect in the reactivation of Atoh1 in the organ of Corti.

1.2 Zusammenfassung

Hörverlust ist die häufigste sensorische Erkrankung in entwickelten Ländern und mehr als 5% der Population in den industrialisierten Ländern haben erhebliche Hörstörungen. In den meisten

Fällen ist die Ursache auf die Degeneration und Zerstörung von Haarzellen und den zugehörigen

Spiralganglienneuronen zurückzuführen. Nur wenige Studien haben über Regeneration des

Gehörsystems berichtet. Es bestehen deutliche Wissenslücken über die molekularen

Interaktionen der Hörfunktion, einschliesslich der erforderlichen Faktoren für die zelluläre

Regeneration und Regulierung der Genexpression in der Cochlea. Daher bleibt Taubheit eine der am weitest verbreiteten und schlecht verstandenen Behinderungen in der Welt, obwohl es einen massiven sozialen und wirtschaftlichen Bedarf an therapeutischen Behandlungen für

Gehörschäden gibt.

Das Ziel der vorliegenden Arbeit ist die Identifizierung von neuen genetischen Faktoren die eine

Rolle bei der Differenzierung und Regenerierung der Haarzellen spielen, um relevante Targets und molekulare Signalwege zu entdecken, welche neue Behandlungsmöglichkeiten für

Hörverlust eröffnen können.

Um neue genetische Faktoren zu identifizieren, die bei der Differenzierung der Haarzellen beteiligt sind, wurde eine Vorläuferzelllinie UB/OC‐1 auf miRNA‐Expression während der

Differenzierung zu Haarzellen untersucht. Durch die funktionelle Charakterisierung von mehreren miRNAs die durch Next‐Generation Sequencing identifiziert wurden, wurde der

Kandidat miR‐210 entdeckt, dessen Runterregulierung die Differenzierung aus einem

Vorläuferzellstadium zu einem differenzierteren Haarzell –Phänotyp führt. Dass die Inhibierung von miR‐210in der UB/OC‐1‐Zelllinie das Zellstadium von proliferierenden zu differenzierten

Zellen verändert, weist darauf hin, dass miR‐210 eine aktive Rolle spielt bei der

Aufrechterhaltung der Vorläuferzellproliferationsphase. Um diese Hypothese zu evaluieren, wurde miR‐210 via Adenovirus in die Maus Corti Organ explants transduziert. Ektopische

Überexpression von miR‐210 in den Corti Organ explants führte zu neuer Haarzellbildung. Durch ein cell lineage tracing Modell wurde die Bildung der neuen Haarzellen aus den ehemaligen Sox2 positiven Epithelialzellen gezeigt. Um die Mechanismen dieser neuen Haarzellbildung, welche von der miR‐210 Expression induziert wurde, besser zu verstehen, wurde eine

Targetsvorhersageanalyse mit Targetscan durchgeführt und mit, in der Literatur veröffentlichten, miR‐210 Genetargets verglichen. Unsere integrative Analyse ergab neue

Kandidaten die mit Immunpräzipitation vom RISC‐Komplex, gefolgt von quantitativer PCR‐

Analyse, validiert wurden.

Neben der Identifizierung mehrerer Transkripte, die bisher noch nicht mit der

Haarzellendifferenzierung in Verbindung gebracht wurden, wurden mehrere von miR‐210 regulierte Transkripte identifiziert, die bekanntlich eine Rolle bei diesem Mechanismus spielen.

Die Daten deuten darauf hin, dass miR‐210 einen neuen möglichen Ansatz für eine Therapie gegen Hörverlust bieten könnte.

Ausserdem wurden Signalwege im Innenohr identifiziert, welche von miR‐210 reguliert werden und einen Angriffspunkt für neuartige Behandlungen von Hörverlust sein könnten.

Im zweiten Teil der vorliegenden Arbeit wurde ein genetisches Screening durchgeführt um

Atoh1 Modulatoren zu identifizieren. Atoh1 ist ein Transkriptionsfaktor der bei der Entwicklung und Regeneration von Haarzellen im Organ of Corti beteiligt ist. Eine kleine Bibliothek von 32 miRNA exprimierenden Plasmiden wurde gescreent. Hierfür wurde eine humane intestinale epitheliale Zellinie (SW480) verwendet, die mit einem Luciferase‐Reporter unter der Kontrolle des regulatorischen Atoh1 Locus konstruiert wurde. Dies führte zur Identifizierung von miR‐152, das die Reaktivierung von Atoh1 Genexpression bewirkt. Reaktivierung von Atoh1 Expression ist mit einer verminderten DNA‐Methylierung der Enhancer‐Region im Atoh1 Locus verbunden.

Diese Ergebnisse stehen im Einklang mit der Hypothese, dass miR‐152 die DNA‐

Methyltransferasen (DNMTs) inhibiert. Weiter wurde der Zusammenhang zwischen miR‐152 und die Atoh1 Aktivierung im Corti Organ charakterisiert. MiR‐152 wurde mit einem viralen

Vektor im Corti Organ überexprimiert. Dadurch wurde keine Atoh1 Aktivierung im Corti‐Organ beobachtet. Diese Ergebnisse zeigen, dass miR‐152 in der Darmkrebs‐Zelllinie über eine epigenetische Modulation Atoh1 aktiviert aber im Corti‐Organ keine Wirkung bei der

Reaktivierung von Atoh1 hat.

2. Introduction

2.1 Hearing loss: Prevalence, causes and medical need

Over 5% of the world’s population – 360 million people – suffer from disabling hearing loss (328 million adults and 32 million children). Approximately one‐third of people over 65 years of age are affected by disabling hearing loss (WHO, World Health Organization, Feb 2014). Individuals with hearing loss carry heavy social, economic, and quality of life burdens. Exclusion from communication can have a significant impact on everyday life, causing feelings of loneliness, isolation and frustration, particularly among older people [1].

The causes of hearing loss can be divided into congenital or and acquired causes. Congenital causes of deafness may be manifested either at the time of birth or may lead to hearing soon after birth. Hearing loss can be caused by hereditary and non‐hereditary genetic factors when genes necessary for the normal development of sensory hair cells are mutated. Mutations in the gene for connexin 26, GJB2, are the most common cause of congenital hearing loss [2]. Such mutations result in developmental malformations of the organ of Corti which contain the sensory hair cells. Hearing loss may also be caused by complications during pregnancy and birth such as maternal infection (taxoplasmosis, rubella, cytomelagovirus), prematurity, post‐natal viral or bacterial infections, or inappropriate use of ototoxic drugs [3, 4] [5] [6].

Acquired causes lead to hearing loss at any age. Infectious diseases such as meningitis, measles and mumps can lead to hearing loss, mostly in childhood, but also later in life. Ear infection such as otitis media can also lead to hearing loss [7]. Excessive noise, including working with noisy machinery, and exposure to loud music or other loud noises can harm a person’s hearing

[8]. Use of ototoxic drugs at any age, such as aminoglycosides, cytotoxic drugs, antimalarial drugs and diuretics, can damage the sensory cells of the inner ear [9] [10]. Aging is also one cause for hearing loss. Age‐related hearing loss (presbycusis) is the gradually loss of hearing that occurs as people get older [6].

Hearing loss occurs when sound waves from the environment are unable to sufficiently stimulate the sensory hair cells of the organ of Corti to send neural impulses to the brain

(conductive hearing loss) or if the sensory hair cells are actually lost (sensorineural hearing loss).

Conductive and sensorineural hearing losses (SNHL) both affect the stimulation of organ of Corti hair cells, but in different ways. In cases of conductive hearing loss, sound waves are hindered from reaching the hair cells in the inner ear, often through increased resistance in the outer ear, tympanic membrane or the middle ear. Because of the resulting decreased amplitude of sound waves reaching the inner ear, conductive hearing loss can be partially overcome with either hearing aids, an auditory prosthesis in which sound waves are amplified prior to entering the ear canal [11]. In SNHL, sound waves enter the ear canal and are transformed into physical vibrations of normal amplitude; however, damaged hair cells are unable to respond to the vibrations and fail to stimulate ganglion neurons [12]. Cochlear implant is a viable option for patients suffering from SNHL, which consists of an electrode array that is placed by surgery into the scala tympani. The electrode array directly stimulates auditory neurons. However current

10 treatments with cochlear implants require an invasive surgical procedure and the efficacy differs greatly among patients. Current treatments for SNHL are not perfect, leaving regenerative medicine as a logical avenue to a perfect cure.

2.2 Anatomy of the inner ear and the organ of Corti

The ear is one of the main sensory organs of the head and is responsible for the senses of hearing, balance and detection of acceleration. The human ear is composed by three anatomical parts: The outer ear is composed of the pinna, or auricle, which is visible projection of the outer ear. This leads to the outside opening of the ear canal which funnels sound waves to the eardrum of the middle ear. The middle ear than transforms sound waves, through the auditory ossicles, into mechanical vibration of the endolymphatic fluid contained in the inner ear and finally the inner ear which is responsible for the transduction of mechanical stimuli into electrical impulses and their propagation to the brain (Figure 1A) [16].

The inner ear is made up of two parts: the vestibular portion, which contains the sensory organs responsible for the senses of motion and position, and the cochlear or auditory region, which contains the sensory organ of Corti responsible for the sense of hearing.

The organ of Corti, which is found within the cochlea contains a single row of approximately

3,500 inner hair cells (IHCs), the primary auditory transducers responsible for conversion of sound into neural signals. The IHC row is separated from the three rows of outer hair cells

(OHCs) by highly specialized pillar cells that enclose the tunnel of Corti, which delineates a mechanical hinge whose movement is important for proper stimulation of the IHC stereocilia.

The OHCs are situated on top of phalangeal supporting cells called Deiters cells. The OHCs play a

11 key role in the frequency‐specific amplification of basilar membrane motion, they connect with their stereociliary bundles to the tectorial membrane that covers the organ of Corti along its entire length (Figure 1B) [13]. Supporting cells are non‐sensory cells that vary greatly in morphological and functional specialization. Supporting cells maintain the correct ionic environment for the hair cells to be functional. They release factors that maintain the trophism and survival of the hair cells [14]. In non‐mammalian species, supporting cells are also capable of serving as progenitors to regenerate hair‐cells after injury [15].

Figure 1 Anatomy of the inner ear

A. The human head illustrating the three (outer, middle and inner) regions of the auditory system(Picture modified from [16]). B. Left: Cross section of the cochlear duct of the snail‐shaped cochlea. The organ of Corti is illustrated with the tectorial membrane above (Picture from:medicalanatomy.net).

Right: Enlarged, a three‐dimensional scheme of the organ of Corti, illustrating the organization of hair cells and supporting cells into rows. Outer hair cells (O1‐3), inner hair cells (IHC), phalangeal supporting cells (PhC), inner pillar cells (IPC), outer pillar cells (OPC)(Picture modified from [16]).

2.3 Molecular mechanisms of inner ear hair cells development and regeneration

2.3.1 Sensory hair cell development

In the mouse, outgrowth of the cochlear duct first becomes evident at approximately E11.

Before E12.5, all of the epithelial cells that comprise the floor of the duct have a similar morphology. Notch 1, Jagged1 , SOX2 are expressed in all cells within the cochlear sensory epithelium [17]. Around E12.5, prosensory cells begin to express p27kip1, a cell cycle inhibitor.

Expression of p27kip1 leads to a termination of mitosis within the prosensory domain [18].

Therefore, the prosensory cells of mammals have no ability to proliferate. Once developing prosensory progenitor cells have become postmitotic, the next step in the development of the sensory epithelia is the specification of a subset of cells to develop as hair cells or supporting cells. The factors that specify cells to develop as either hair cells or supporting cells are still poorly understood. Howerever, following terminal mitosis, some prosensory cells begin to express Atoh1.

The basic helix‐loop‐helix (bHLH) transcription factor Atoh1 (also knowns as Math1) is the earliest known gene expressed in the prosensory domain eventually associated with the development of sensory hair cells [19]. Atoh1 is expressed prior to all other known hair cell

13 genes, and is later down‐regulated in maturing hair cells [20]. In the embryonic Atoh1‐null mice, the progenitors for hair cells die at about the time when they would normally differentiate into hair cells, and thereby failed to generate cochlear hair cells [21]. In contrast, the overexpression of the Atoh1 gene in the cochlear non‐sensory region generated hair cells [22].

Although the targets of Atoh1 in developing hair cells are largely unknown, Atoh1 is known to bind to its own enhancer region to initiate a positive‐feedback loop that will act to maintain or increase its own expression [22]. Studies have demonstrated that Six1 and its transcriptional co‐activator Eya1, which are expressed in the prosensory region of the cochlea, can bind directly to the Atoh1 autoregulatory enhancer [23].

Several factors that positively or negatively regulate Atoh1 have been identified. Studies in cell lines have suggested that the canonical Wnt signaling pathway can act to induce Atoh1 expression as a result of direct binding to a Tcf/ Lef‐binding site within the Atoh1 enhancer region [24]. Although not confirmed in the inner ear, these results would suggest that Wnt signaling acts as a positive regulator of Atoh1 expression. The best characterized pathway involved in the regulation of hair cell development and Atoh1 expression is the Notch signaling pathway. Differentiating hair cells are able to actively suppress hair cell fate in their supporting cell neighbors through the Notch signaling pathway [16]. Notch receptors and their ligands are both transmembrane proteins and therefore can only bind in cells that are directly in contact with one another. In the inner ear, localization studies have shown expression of Notch1 throughout the epithelium and expression of two Notch ligands, Jagged2 and Delta‐like1, in developing hair cells [25]. Moreover, several bHLHs, including Hes5 and Hes1, are expressed in developing supporting cells and have been shown to inhibit Atoh1 expression [26]. Deletion of

14 different members of the pathway results in varying increases in the number of hair cells, an effect that is very consistent with classic Notch‐mediated lateral inhibition [27](Figure 2). The role of Notch signaling in maintaining hair cell and supporting cell fate during cochlear development has led the speculation and demonstration that inactivation of Notch signaling promote hair cell regeneration in the damaged mammalian cochlea [28].

Understanding the factors that regulate the development of the sensory hair cells is important, not only to increase the understanding of ear development and its functional physiology but also to shed light on how these cells may be replaced therapeutically.

Figure 2: Scheme for how Notch signaling maintains hair cell and supporting cell fates (Picture from [27]).When Notch signaling is blocked in supporting cells, Atoh1 is no longer repressed and supporting cells trans‐differentiate into hair cells (bottom left). In Atoh1 based gene therapy, the ectopic expression of Atoh1 leads to drive supporting cells to hair cell fate.

2.3.2 Mechanisms of hair cell regeneration

In contrast to the permanent nature of mammalian deafness, all non‐mammalian vertebrates have the capacity to regenerate their hair cells in the organ of Corti. In birds for example, damage to the hair cells of the organ of Corti brings about a re‐initiation of proliferation of supporting cells [29, 30]. There are two theories to explain how these newly proliferating supporting cells replace lost hair cells. Either cell cycle re‐entry of the supporting cell causes their dedifferentiation allowing the supporting cell to form a neurosensory precursor cell. This precursor cell can then undergo further cell division to form two cells with subsequent differentiation into one hair cell and one supporting cell. The alternative hypothesis is that cell cycle re‐entry of the supporting cell forms two supporting cells (with no dedifferentiation), with one of the supporting cells transdifferentiating into a hair cell [27](Figure 3).

Supporting cells in mammals do not normally divide or transdifferentiate when hair cells are lost, and so regeneration does not occur. To understand the failure of mammalian hair cell regeneration, there is a need to understand the molecular mechanisms that underlie cell division control and hair cell differentiation during embryogenesis and in postnatal mouse.

The cyclin‐dependent kinases (CDK) and their regulatory machinery are prime targets of developmental signaling to co‐ordinate proliferation with differentiation. During development, pro‐sensory progenitor cells in the organ of Corti proliferate until expression of the cell cycle inhibitor p27Kip1 induces a cell‐cycle arrest and terminal differentiation. During the

16 differentiation process, p27Kip1 is rapidly downregulated in hair cells, but is maintained at high levels in differentiating supporting cells [31, 32]. This suggests the hypothesis that changing levels of p27kip may be involved in the maintenance of the postmitotic state of hair cells and supporting cells. In p27 deficient knock‐out mice, cell division in the organ of Corti continues past embryonic day 14 when mitosis is normally completed, leading to supernumerary production of supporting and hair cells [33]. The resulting organization of the hair cell area is incompatible with a normally active cochlea and leads to severely impaired auditory function in p27‐/‐ mice. From a therapeutic perspective, although analysis of isolated postnatal supporting cells in vitro suggested that p27Kip1 represents a suitable target for hair‐cell regeneration, induced p27kip1 depletion ex vivo and in vivo revealed no trans‐differentiation of supporting cells into hair cells and identified obstacles that need to be overcome in order to achieve hair cell regeneration via stimulation of supporting cell proliferation in vivo [34].

Transdifferentiation is the process of directly transforming from one differentiated cell type into a unique second differentiated cell type. When hair cells are lost in birds, supporting cells can undergo direct transdifferentiation into hair cells in response to the loss of hair cells. This is marked by the upregulation of the transcription factor Atoh1 in the differentiating cells [35].

Given that Atoh1 is the earliest known marker for hair cell differentiation, both in embryogenesis and avian regeneration, understanding its regulation may be key stimulating regeneration in mammals. Indeed, forced expression of Atoh1 in prenatal rodent supporting cells causes the differentiation of these cells into sensory hair cells, both in vitro and in vivo [36,

37].

Recent studies showed that using adenovirus to express Atoh1 in the organ of Corti of mature guinea pigs that had been deafened by administration of ototoxic drugs, had a significant recovery of hair cells in parts of the cochlea, together with some improvement in auditory brainstem [38].

Figure 3: Diagram illustrating two different modes of avian hair cell regeneration.

After a traumatic injury to avian sensory epithelium, supporting cells can replace the dead hair cells either bydirectly transdifferentiating into hair cells (bottom left), or by re‐entering the cell cycle and undergoing asymmetric differentiation, where one daughter cell trans‐differentiates into a hair cell and the other remains as a supporting cell (bottom right) (Picture from [27]).

2.5 The importance of microRNAs in inner ear hair cell development

MiRNAs are conserved small non‐coding RNA molecules that have crucial roles in regulating gene expression and cell fate. MiRNAs regulate cell physiology by fine‐tuning the tissue and cell‐ type specific expression of multiple target RNAs through several post‐transcriptional mechanisms like inhibition of translation and induction of mRNA destabilization and decay [39].

In mammals, miRNA genes are usually transcribed by RNA polymerase II promoters and then processed into mature miRNAs through canonical or non‐canonical miRNA biogenesis pathways.

During canonical miRNA biogenesis, the primary miRNA (pri‐miRNA) hairpin is digested to precursor miRNA (pre‐miRNA) by Drosha, a member of the RNase III family. Non‐canonical miRNA biogenesis differs at this step in that pre‐miRNAs are generated by the mRNA splicing machinery, circumventing the requirement for Drosha‐mediated digestion in the nucleus.

In both pathways, the pre‐miRNAs are exported to the cytoplasm via the nuclear export protein exportin 5 and further processed by a second RNase III enzyme, Dicer. The mature double‐ stranded miRNAs are then loaded into a functional ribonucleoprotein complex called the RNA‐ induced silencing complex (RISC), which serves as the catalytic engine for miRNA‐mediated posttranscriptional gene silencing. RISC consists of multiple protein factors, including Argonaute proteins which are the key catalytic enzymes within the complex. Argonaute proteins bind miRNAs and are essential for their downstream gene‐regulatory mechanisms to regulate mRNA degradation and protein expression [39].

MiRNAs are associated with many developmental processes and diseases and have been found to be crucial for normal development in animals [40].

The first description of miRNAs being involved in the development of the inner ear came with the description of miR‐9 as a potential regulator of COL9A1 [41]. This was followed by microarray analyses of the developing inner ear, where a conserved cluster of miRNAs (miR‐96, ‐

182 and ‐183) was found to have their expression restricted to the inner ear [42]. In situ hybridization revealed the unique expression pattern of mir‐182, mir‐183 and mir‐96 in inner and outer hair cells of the cochlea, hair cells of the vestibular organs and spiral and vestibular ganglia [43].

From that time on, the field exploded with a series of discoveries, the most compelling being the identification of a mutation responsible for deafness in two extended Spanish families. The causative mutation lay in the seed region of miR‐96. This was the first example of a mutation found in a miRNA to lead to a Mendelian disease, and the first miR mutation shown to cause deafness [44]. MiRNAs were also described in other ear‐related diseases such as the role of miR‐

21 in human cholesteatoma growth and proliferation and in vestibular schwannomas [45].

Friedman et al. [46] showed that Dicer activity is required for the survival of normal and functional hair cells in the cochlear sensory epithelia. Removal of Dicer, which initially does not perturb development, causes abnormal growth and subsequent degeneration of hair cells, leading to deafness. Therefore, miRNAs have an essential role in inner ear development and function.

In a different model system, zebrafish were used to understand the role of the miR‐96, miR‐182 and miR‐183 cluster in inner ear regeneration. The cluster was overexpressed in zebrafish by injecting embryos with double‐strand miRNAs. Embryos with miR‐182 and miR‐96 over‐

20 expression exhibited body malformations and produced ectopic hair cells [47]. A knockdown of each of the miRNAs in the cluster, using morpholino oligonucleotides, showed a decrease in hair cell numbers. Overexpression of miR‐182 in a miR‐96 knockdown embryo demonstrated a partial rescue effect, and the number of hair cells was increased as compared to the knockdown[47].

These findings and additional studies in following years have led to the idea that miRNAs may have a promising therapeutic aspect for sensory hair cell regeneration, either as being the active agent promoting regeneration or as mediator helping to uncover downstream targets involved in regeneration.

2.6 In vitro and ex‐vivo models for assessment of hair cell differentiation

2.6.1 The inner ear cell line UB/OC‐1

Conditionally immortal cell lines from mammalian sensory epithelia provide important experimental tools in the study of the development, function and regeneration of hearing [48].

UB/OC‐1 is a conditionally immortal cell line derived from the organ of Corti of the

Immortomouse embryo [49]. The Immortomouse carries a transgene that expresses the thermolabile simian virus 40 (SV40) large T antigen tsA58 under the control of the Interferon‐ inducible murine H‐2Kb promoter [50]. UB/OC‐1 cells were selected from the Immortomouse at embryonic day 13 (E13), to represent the developmental stage of hair cell progenitors. At this

21 stage, the cells have not fully differentiated into sensory hair cells and retain the capacity to further differentiate into hair cells (Figure 4)[49].

The cell line UB/OC‐1 can be induced to differentiate in a conditional manner. The UB/OC1 cell line expresses the immortalizing transgene tsA58, regulated by an interferon inducible promoter. When the cells are cultured at 33°C in the presence of interferon‐gamma, proliferation is maintained, but following removal of interferon‐gamma from the culture media and an increase in temperature to 39°C, proliferation ceases and the cells start to differentiate into hair cells (Figure 5) expressing the POU‐domain transcription factor Pou4f3 (Brn3.1).

Interestingly, UB‐OC1 cells do not express Atoh1, a master transcription factor for hair cell differentiation, leading to the hypothesis that alternative, however yet unknown, pathways might exist driving this process [51].

Investigation in the hair cell precursors permit functional studies of cochlear genes and allow the exploration of key factors and signaling pathways that may facilitate therapeutic stimulation of hair cell replacement.

Figure 4: Origin of the embryonic UB/OC‐1 cell line.

A pluripotent progenitor passes through its last mitosis at approximately E14, giving rise to a non‐ sensory and a prosensory precursor. The prosensory precursor has the potential to differentiate without undergoing mitosis into either a committed hair cell precursor or a committed supporting cell precursor.

UB/OC‐1 has been immortalized prior to the expression of Pou4f3 (Brn3.1) UB/OC‐1 represents the uncommitted prosensory precursor with the potential to differentiate into hair cells [49].

Figure 5: UB‐OC1 cells at 33°C reflect the undifferentiated precursor phenotype. The cells differentiate towards becoming more “hair cells‐like” expressing Pou4f3 and changing morphology, after exposure to higher temperatures (39°C) and the removal of Interferon‐γ.

2.6.2 Ex‐vivo assay with organ of Corti explants

Mammalian inner ear sensory epithelia (organ of Corti) is a tiny delicate tissue embedded in a bony labyrinth of the temporal bone and are not readily accessible for non‐traumatic, nonsurgical interventions. An increasingly essential tool in auditory research is the isolation and in vitro culture of organ of Corti explants. Once isolated, the morphology and the molecular characteristics of sensory hair cells and non‐sensory supporting cells within the organ of Corti explant cultures resemble those observed in vivo and can be studied within its intrinsic cellular environment. The explants may be utilized in several ways to provide information regarding gene expression, stereocilia motility, cell and molecular biology, as well as the assessment of hair cell regeneration. Another important use of cultivated explants is the development of therapeutics, such as compound treatment or investigation of the effect of gene

24 overexpression. The organ of Corti from mutant mice that die before birth can be cultured so that their in vitro development and responses to different factors can be analyzed. The limitations of this ex vivo model are: the number of organs of Corti that can be isolated, the isolation of the organ of Corti from adult inner ear and the delivery of exogenous DNAs to the cells of the organ. The inner ear hair cells cannot be transfected using conventional transfection techniques such as lipofection or electroporation, because of terminally differentiated cells.

However, viral vectors such as Adenovirus and Adeno associated virus, are powerful tools for gene delivery to the organ of Corti in vitro and in vivo. One of the major advantages of viral vectors is their good efficiency of transfection and distinct tropism for different cell types, which can help restrict transgene expression to a subset of cells.

Figure 6: Organ of Corti explants as ex‐vivo tool for hair cell differentiation assessment. Organ of Corti explants were isolated from the inner ear of a hair cell reporter mouse. The explants are taken in culture and treated with compounds or transduced with viral vectors for the analysis of hair cell differentiation.

3 Aim and outline of this dissertion

The aim of the present work was to identify novel genetic factors playing a role in the differentiation and regeneration of inner ear hair cells, in order to discover relevant targets and molecular pathways that might open up new possibilities for the treatment of hearing loss.

Two different approaches were used to identify factors playing a role in hair cell differentiation and regeneration.

First, a genome‐wide mRNA and non‐coding RNA sequencing of an embryonic inner ear cell line

(UB/OC1) was conducted, using next generation sequencing to identify genes playing a role in sensory hair cell differentiation. This sequencing experiment showed that miR210 was upregulated in the precursor cell stage of UB/OC1. This observation was further validated using a functional hair cell differentiation assay in vitro. Adenovirus‐mediated delivery to cochlear explants was used to validate the capacity of miR210 expression to promote hair cell differentiation. Overexpression of miR210 in mammalian cochlear explants promoted ectopic hair cell formation in the greater epithelial ridge. Using lineage tracing, it was found that these new hair cells arise from Sox2‐expressing supporting cells. To further explore this observation, potential miR210 target genes were predicted and validated using a pull‐down experiment that monitored the target genes using a PCR‐array. This experiment highlighted genes in a number of pathways that could be controlling inner ear differentiation (Chapter 4).

Second, a genetic screen was used to identify genetic elements that ‐ when modulated ‐ lead to reactivation of Atoh1 gene expression. Atoh1 is a key transcription factor involved in the differentiation and regeneration of sensory hair cells of the inner ear. For this purpose, a

26 reporter cell line was engineered with a luciferase reporter gene under the control of the entire

Atoh1 regulatory locus, so that transcriptional activation leads to increased luciferase expression. As part of these efforts, a small set of plasmids expressing micro‐RNAs was tested.

One of the actives identified from this screening was further characterized for its molecular mechanisms and epigenetic reactivation of Atoh1 (Chapter 5).

4 Identification and characterization of novel genetic factors involved in the regeneration of inner ear hair cells

4.1 Mir‐210 promotes ectopic sensory hair cell formation via transdifferentiation of supporting cells in the organ of Corti

4.1.1 Abstract

Hearing loss is the most common sensory defect, with several hundred millions of people worldwide having hearing disorders. In most cases, the cause of hearing loss is related to the degeneration and death of hair cells and their associated spiral ganglion neurons. Despite a growing body of knowledge about the sources of these disorders, relatively few studies have reported regeneration of the auditory system. Significant gaps remain in our understanding of the molecular mechanisms underpinning auditory function, including the factors required for sensory cell regeneration. Recently, the identification of transcriptional activators and repressors of hair cell fate has been augmented by the discovery of microRNAs (miRNAs) associated with hearing loss. Identification of miRNAs and their gene targets may reveal new pathways for hair cell regeneration and thereby provide new avenues for the treatment of hearing loss.

In order to identify new genetic elements enabling regeneration of sensory hair cells in the inner ear, next‐generation miRNA sequencing (miR‐Seq) was used to identify the most prominent microRNAs expressed in the mouse embryonic inner ear cell line UB‐OC1. Based on miR‐Seq, we selected the eight most highly represented miRNAs for further characterization.

MiR‐210 knock‐down in vitro resulted in hair cell marker expression in UB‐OC1 cells, whereas

28 ectopic expression of miR‐210 resulted in new hair cell formation in cochlear explants. By using lineage tracing mouse models, we identified trans‐differentiation of supporting cells as the mechanism for new hair cell formation. Potential miR‐210 targets were predicted in silico and identified experimentally using a miR‐trap approach.

MiR‐Seq, followed by ex vivo validation, revealed miR‐210 as a novel factor driving trans‐ differentiation of supporting epithelial cells into sensory hair cells. Our data suggest that miR‐

210 might be a potential new factor for hearing loss therapy. Moreover, identification of pathways regulated by miR‐210 in the inner ear could reveal novel drug targets for the treatment of hearing loss.

4.1.2 Introduction

Sensorineural hearing loss is the most common sensory deficit in the world, accounting for more than 300 million people worldwide [52]. In most cases, the cause of hearing loss is related to the degeneration and death of hair cells and their associated spiral ganglion neurons [53], where damage is due to acoustic over‐stimulation, infection, ototoxic drugs and aging. About

4% of individuals under 45 years and about 34% of those over 65 years have debilitating hearing loss. Importantly, there is a strong indication that hearing impairment is becoming more common among young adults and children, particularly due to new trends like exposure to portable music players [6]. Despite the need, patients have few options: There are cochlear implants but currently no drugs for hearing loss [52], and although there is a massive social and

29 economic demand to develop therapeutic treatments for hearing loss, deafness remains one of the most widespread, costly and poorly understood disabilities in the world.

Besides increasing knowledge in tissue regeneration, significant gaps remain in our knowledge regarding the molecular interactions underpinning auditory function, including the factors required for cellular regeneration and regulation of cochlear gene expression. Whereas non‐ mammalian vertebrates can replace hair cells through transdifferentiation of epithelial supporting cells [54, 55] [54, 56], this spontaneous regenerative capacity has been lost in mammals. However, in mammals, supporting cells can be forced to transdifferentiate into new hair cells given the right stimulus, namely over‐expression of Atoh1, which is normally only present during fetal development [57]. This suggests that the molecular activity required for inducing hair cell fate is still present and functional in adult mammalian supporting cells, and that supporting cell fate may be altered if the cells receive the appropriate signals [58, 59].

Alternative factors for sensory hair cell regeneration were recently described, such as cell cycle genes like cyclin‐dependent kinase inhibitor 1B (p27Kip1). During development, pro‐sensory progenitors in the organ of Corti proliferate until expression of cell cycle inhibitor p27Kip1 induces cell cycle arrest and terminal differentiation [60]. In p27 deficient knock‐out mice, cell division in the organ of Corti continues past embryonic day 14 when mitosis is normally completed, leading to a supernumerary production of supporting and hair cells in the organ of

Corti [60] [33]. However, the organization of the hair cell area is incompatible with a normally active cochlea and as a result, auditory function is severely impaired in p27 deficient mice [33].

From a therapeutic perspective, although analysis of isolated postnatal supporting cells in vitro suggested that p27Kip1 represents a suitable target for hair‐cell regeneration, induced p27kip1

30 depletion ex vivo and in vivo revealed no trans‐differentiation of supporting cells into hair cells and identified obstacles that need to be overcome in order to achieve hair cell regeneration via stimulation of supporting cell proliferation in vivo [34] [61] [62].

MicroRNAs are conserved small non‐coding RNA molecules that have crucial roles in regulating gene expression and cell fate. MicroRNAs regulate cell physiology by fine‐tuning tissue‐ and cell type‐specific expression of multiple target RNAs through several post‐transcriptional mechanisms, including inhibition of translation [63] and induction of mRNA destabilization and decay [64]. By controlling multiple target RNAs simultaneously, some miRNAs were found to modulate several components of a single pathway whereas others were found to modulate biological processes by e.g. targeting distinct RNAs in key cell proliferation pathways [65].

Recently, the importance of miRNAs in inner ear development and their role in maintenance of hearing has been demonstrated in multiple animal studies including zebrafish and rodents [42,

46, 66]. Moreover, the essential role of miRNAs in auditory function became evident by the discovery of mutations in miR‐96 which underlie non‐syndromic hearing loss in humans [44].

These findings and additional studies in the following years have led to the idea that miRNAs may represent a promising therapeutic aspect for sensory hair cell regeneration, either as being the active agents promoting regeneration or as mediators helping to uncover downstream targets involved in regeneration [67].

Next‐generation RNA sequencing (RNA‐seq) has brought remarkable opportunities for the discovery of differential gene expression including miRNAs. While RNA‐seq has been widely used in multiple fields to identify and characterize miRNAs, the technology has just started to be

31 exploited for miRNA profiling in the inner ear where RNA‐seq revealed a number of miRNAs being differentially expressed between cochlear and vestibular sensory epithelia [68]. To identify miRNAs differentially expressed during hair cell differentiation, we conducted genome‐ wide next generation non‐coding RNA sequencing of the inner ear cell line UB/OC‐1, a conditionally immortalized cell line derived from a population of non‐sensory epithelial cells in the greater epithelial ridge (GER) that has the potential to differentiate into a hair‐cell‐like phenotype, without the intervention of Atoh1 [49, 51]. UB/OC‐1 cells were previously shown to resemble proliferative progenitors and, based on their capability of differentiating into hair cells, provide a useful tool for studies on gene expression profiling and mechanisms of mammalian cochlear hair cell differentiation/regeneration [69].

Non‐coding RNA sequencing identified several miRNAs being differentially expressed during

UB/OC‐1 cell differentiation. Functional validation of the most prominent down‐regulated miRNAs revealed that depletion of miR‐210 triggers differentiation of UB/OC‐1 cells towards the hair cell‐like phenotype, indicating a possible role in maintaining the proliferative progenitor state. To explore whether the reciprocal approach can force non‐sensory epithelial cells to switch to the hair cell phenotype, we overexpressed miR‐210 in cochlear explants and identified ectopic hair cell formation in the greater epithelial ridge. Using lineage tracing, we confirmed that newly formed hair cells arise from Sox2‐expressing supporting cells. To further explore the mechanisms of miR‐210 function, potential miR‐210 target genes were predicted using different algorithms and ultimately functionally validated using a miR‐210 pull‐down assay. Our experiments identify miR‐210 as a new factor with the potential to drive non‐sensory epithelial

32 cells towards a sensory hair cell phenotype and identify putative downstream targets mediating this effect.

4.1.3 Material and Methods

Cell culture and RNA isolation

UB/OC‐1 cells were kindly provided by Prof. Matthew Holley (Institute of Molecular Physiology,

Sheffield, UK) and cultured in Minimal Essential Medium with Earle's salts and Glutamax I (Life

Technologies), 10% fetal calf serum (Life Technologies), and 50 U/mL of mouse γ‐interferon (γ‐

IFN, Life Technologies) at 33°C, 5% CO2. Differentiation of UB/OC‐1 cells was induced by removing γ‐IFN from the growth medium and incubation at 39°C, 5% CO2 as previously described [49]. RNA samples were prepared from cells at day 0 (before differentiation) and 24 hours after differentiation was induced. Experiments were performed in triplicate. Total RNA was prepared using the DirectZol Kit (Zymoresearch) according to the manufacturer’s instructions. All RNA samples had an RNA Integrity Number (RIN) of 8.5 or higher.

RNA‐Sequencing

Small RNA libraries were generated using the Illumina TruSeq Small RNA Sample Preparation

Kits. Sequencing was performed in single end mode, 1x50 bp, on the Illumina HiSeq 2500 platform, following the manufacturer’s protocol. Images from the instrument were processed using the manufacturer’s software to generate FASTQ sequence files. Read quality was assessed by running FastQC (version 0.10) on the FASTQ files. Sequencing reads showed excellent quality, with a mean Phred score higher than 30 for all base positions. A total number of 302 million 50‐

33 bp single reads (e.g. from a minimum of 28,9 million to a maximum of 44,2 million reads per sample) were trimmed using fastx clipper [FASTX‐Toolkit, http://hannonlab.cshl.edu] to remove remnants of the 3’‐adapter sequence. Trimmed reads were aligned to the Mus musculus miRBase v. 19 hairpin reference sequences [70] using the Bowtie short‐read aligner [71]. The percentage of trimmed reads mapping to the miRNA hairpin sequences varied from 47,3 % to

67,5 %. The microRNA abundance was quantified using an in‐house NGS analysis pipeline, counting aligned reads for each microRNA that intersect the mature sequence region of the hairpin (overhang of 4 bp allowed), normalized by the individual sequencing library sizes.

Summary alignment statistics are shown in Table 1. Differential miRNA expression analysis between the precursor and differentiated UB/OC‐1 cells was performed using the

R/Bioconductor package DeSEQ, considering a cut‐off of at least 2 fold change in expression and an adjusted p‐value <0.01 [72].

Table 1: Number of miRSeq reads and alignment statistics

Locked Nucleic Acids transfection and RT‐PCR

UB/OC‐1 cells were transfected with 50 nM LNA (mircury, Exiqon) using Lipofectamine RNAimax

(Life Technologies) according to the manufacturer’s recommendations. Seventytwo hours post‐ transfection, RNA was extracted using Trizol and the Directzol extraction kit (Zymoresearch). cDNA was prepared using High Capacity Polymerase (Applied Biosystems). Primers employed for the detection of the transcript Pou4f3 and GAPDH are:

GAPDH, position 248 (5'AACGGGAAGCCCATCACC 3') and 672 (5' CAGCCTTGGCAGCACCAG 3');

Pou4f3, positions 205(5'CCATGCGCCGAGTTTGTCTCC 3') and

639 (5'CTCCACATCGCTGAGACACGC 3');

Reporter mice

To generate a targeting vector for homologous recombination, Sox2 genomic sequences were amplified from C57Bl/6 mouse genomic DNA and Sox2 homology arms were cloned into a targeting vector containing IRES‐CreERT2 and rox‐flanked neomycin cassettes. Following introduction into C57Bl/6 embryonic stem cells, neomycin resistant clones were screened by polymerase chain reaction (PCR) for homologous recombination. Correct targeting was confirmed by Southern blot using a neomycin‐specific probe that allowed the exclusion of random integration events of the targeting vector. Selected targeted ES cells were injected into

BALB/c blastocysts and chimeric mice were bred with C57Bl/6 females to obtain Sox2‐IRES‐

CreERT2 knock‐in mice. To eliminate the rox‐flanked neomycin cassette, Sox2 gene targeted

35 mice were crossed with a mouse line expressing Dre recombinase and analyzed for the loss of the neomycin cassette.

Lineage Tracing

Sox2‐IRES‐CreERT2 knock‐in mice were crossed to CAG‐floxed EGFP mice to facilitate Sox2 mediated lineage tracing [58]. At postnatal days 0, 1, 2, and 3, 100 μl tamoxifen dissolved in corn oil (50 mg/ml) were injected i.p. into mothers, and tamoxifen was taken up by pubs by suckling. Administration of tamoxifen results in Cre‐mediated excision of the floxed Stop cassette, followed by permanent EGFP expression in Sox2 expressing cells and cell lineages derived thereof [59].

Adenovirus generation

For adenovirus construction, the Virapower Adenoviral Expression system (Life Technologies) was used. Briefly, the genomic region for pre‐mir210 (miRBase MI0000695) was cloned into pENTR and inserted into the pAd/CMV/V5‐DEST plasmid (Life Technologies) by Gateway recombination. The PacI linearized pAd‐miR210 expression construct was transfected into 293A packaging cells (Life Technologies) as recommended by the manufacturer. Virus was collected

10 days after infection and stored at − 80 °C unl needed. Viral ters were measured by standard end‐point dilution assay in 293A cells and miR‐210‐5p expression was confirmed by

Taqman® analysis of transfected 293A cells (Life technologies) (Figure 7).

Figure 7: Fold increase of miR‐210‐5p expression after CMV‐miR210 adenoviral transduction of HEK293A cells.

Organ of Corti explants

Double transgenic pubs from lineage tracing were identified by genotyping, dissected at P4 and the cochlea was removed from the temporal bone. The isolated cochlea was transferred to

Hanks’ balanced salt solution (Life Technologies) and the organ of Corti was isolated as described by Parker et al. [73]. The stria vascularis and basal hook region was removed and the organ of Corti was transferred to ice‐cold HBSS supplemented with 10 mM HEPES buffer (pH

7.3). Next, organs of Corti were transferred to 0.4 cm2 well culture plates (BD Biosciences

Discovery) pre‐coated with CelltakTM (BD Biosciences Discovery), containing 100 μL of

Dulbecco's modified Eagle's medium–F12 (Gibco) and supplemented with 1 μg/mL ampicillin

(Gibco) and 1% FBS (Gibco), and were kept at 37 °C in 5% CO2 for 1 day. For adenovirus transduction, 1x109 virus particles were added and organs of Corti were analyzed 3 days after transduction.

Immunostaining

Explants were fixed in 4% paraformaldehyde. The specimens were washed in PBS and then blocked for 1 hour with blocking buffer (PBS, 5% of donkey serum (Sigma) and 0.1% of Triton X‐

100). Myosin7a rabbit polyclonal antibody (Proteus Biosciences), Sox2 goat polyclonal antibody

(Santa Cruz), GFP polyclonal chicken antibody (Life Technologies), diluted in the blocking buffer were added and incubated overnight at 4°C. The specimens were incubated for one hour at room temperature with the secondary antibody (Life Technologies) in the blocking buffer.

Samples were washed and mounted with FluorsaveTM (Calbiochem).

Cell count analysis

Hair cells were counted in 100 μm segments along the length of the cochlea as described (Lit).

Each group had at least three different cochlear explants and each explant was sampled in five different areas. The cell counts were determined by manually counting the cells in the confocal images. Values are expressed as the mean ± standard error and using a student paired t‐test for statistical analysis, a p <0.05 was considered to indicate a statistically significant difference. miRNA target prediction

TargetScan Mouse 6.2 and Human 6.2 version, along with associated 3'UTR multi‐species alignment supporting files, were downloaded from http://www.targetscan.org/ and run on a

Unix environment with Perl 5.14.1. Three prediction analyses were performed in parallel. First,

TargetScan Mouse v6.2 was used to retrieve predicted targets of mouse mmu‐miR‐210‐3p

(Sanger Accession: MIMAT0000658) and mmu‐miR‐210‐5p (Sanger Accession: MIMAT0017052).

Second, TargetScan Human 6.2 was used to retrieve predicted targets for human hsa‐miR‐210‐

3p (Sanger accession: MIMAT0000267) and hsa‐miR‐210‐5p (Sanger accession:

MIMAT0026475). Third, TargetScan Human 6.2 was also used to retrieve predicted targets for the mouse mmu‐miR‐210‐3p and mmu‐miR‐210‐5p (as the mouse 3'UTRs, genome‐wide, are not as well annotated as those for the human genome, the TargetScan team recommends retrieving target predictions of mouse miRNAs using TargetScan Human in addition to using

TargetScan Mouse; see FAQ section for more details: http://www.targetscan.org/faqs.html).

Target predictions annotated with at least one conserved site were retained to limit the number of predictions to carry forward. The three lists were combined, using mouse‐human homolog gene ID relationships retrieved from Homologene V67.

Metacore/Metabase (version 6.15; Thomson Reuters) was used to retrieve miR‐210 targets reported in peer‐reviewed literature. MiRNA‐mRNA interactions annotated in this repository contain a ‘trust’ field with values from best (Present) to less (Probably Present) to worst

(NLP=Not Likely Present). The latter may contain interactions that were predictions from

TargetScan or other miRNA target prediction programs that were otherwise not further characterized in the corresponding paper. Target candidate lists from TargetScan and Metabase for mmu‐miR‐210‐3p were integrated and prioritized based on the following criteria:

1. Known miRNA‐mRNA interactions reported in Metabase for mouse with a trust level

“present”, pubmed ID available);

2. Present in all 3 TargetScan predictions, miRNA‐mRNA interactions reported in Metabase with trust level ”probably present” or “present”, pubmed ID avalaible;

3. Present in all 3 TargetScan prediction lists, miRNA‐mRNA interactions reported in

Metabase but with trust level “not likely present”, pubmed ID available;

4. Present in all 3 TargetSan predictions, no Metabase/Literature interaction;

5. Prediction for mmu‐miR‐210 is present in TargetScan mouse or human, no

Metabase/Literature interaction.

Pulldown of microRNA/RISC complex and qPCR array

To capture miR‐210 target mRNAs, the MirTrap System (Clontech) was used. UB/OC‐1 cells were co‐transfected with 20 μg of pMirTrap vector (Clontech) and pCMV‐mir210 vector (Origene), using Lipofectamine 2000 (Invitrogen). Forty eight hours post transfection, cells were rinsed with cold PBS and extracted in cold lysis buffer following the MirTrap System protocol. Bead‐ bound RNA was isolated using the NucleoSpin RNA XS kit (Macherey Nagel), and cDNA was synthesized using the RT PreAmp cDNA Synthesis Kit (Qiagen). MiR‐210 targets were identified by qPCR, using the array for predicted miR‐210 mouse targets (PAMM‐6009ZE‐1, Qiagen). Fold enrichment was calculated from Ct values and normalized.

4.1.4 Results

Ninety‐nine miRNAs significantly change their expression during UB‐OC1 differentiation

To investigate the potential role of microRNAs in hair cell formation, we performed next generation small RNA sequencing analysis (miRseq) of UB/OC‐1 cells at the non‐sensory GER precursor stage as well as differentiating cells. The cell line UB/OC‐1 was previously derived from Immortomouse embryos and can be induced to differentiate in a conditional manner.

When cells are cultured at 33°C in the presence of gamma‐interferon, proliferation is maintained, but following removal of gamma‐interferon from the culture media and an increase in temperature to 39°C, proliferation ceases and the cells start to differentiate [49, 51].

For miRseq, we collected five samples of UB/OC‐1 cells at GER stage grown at 33°C and three samples of differentiating UB/OC‐1 cells collected one day after temperature shift to 39°C. On average, 37.7 million reads were sequenced per sample (Table 1) and 47.3% to 67.5% of trimmed reads aligned to mouse miRNAs annotated in miRBase version 19 [74] [75]. All reads were used to detect a total of 687 and 647 distinct mature miRNAs expressed in the precursor and differentiating stages of UB/OC‐1 cells, respectively. MicroRNA counts showed very high correlation between sample replicates (Spearman’s correlation coefficient, R > 0.92). We reduced sample depth‐specific bias by dividing the raw counts by the total number of million aligned reads per sample, i.e. reads per million reads (RPM). Principal component analysis of such normalized microRNA expression counts showed consistency across all replicate samples from the same group (Figure 8). Using these data, we identified 99 mature miRNAs that significantly change expression one day after initiating UB/OC‐1 differentiation (>2 fold change,

FDR adjusted p‐value < 0.01). Of these, expression of 50 miRNAs was enhanced and expression of 49 miRNAs was repressed during early stages of differentiation (Figure 9).

Figure 8: Principal component analysis of five and three replicates from control 33°C (blue) and 39°C treated sample (red) groups. The samples cluster according to sample group.

Figure 9: Differentially expressed microRNAs in UB/OC‐1 cells following differentiation. Heat map representing color‐coded expression levels of differentially expressed microRNAs (up‐ or down‐regulated >2‐fold, FDR adjusted p‐value < 0.01) in UB/OC‐1 at precursor stage (33°C) and 24h after induction of differentiation (39°C). MicroRNAs are ranked by fold change. Colors range from bright pink (low expression) to dark red (high expression). We used a row level scaling to better account for the considerable dynamic range. The row maximum ranges from 3 to about 46’000 RPM (dark red). Expression of 50 miRNAs was enhanced and expression of 49 miRNAs was repressed during early stages of differentiation.

Blockade of miR‐210 induces expression of the Pou4f3 hair cell marker in UB/OC‐1

To evaluate a direct effect of miRNA expression on maintaining the UB/OC‐1 hair cell precursor cell stage, we selected 8 of the most differentially expressed miRNAs with high expression at precursor stage for further analysis (Table 2). For this, UB/OC‐1 cells were transfected with LNA

(locked nucleic acid) miRNA antagonists and expression of the early hair cell marker Pou4f3 [76] was analyzed by RT‐PCR 72 h after transfection. Of the 8 miRNAs investigated, inhibition of miR‐

210 induced a strong increase of the expression of the Pou4f3 hair cell marker in UB/OC‐1 cells

(Figure 10), indicating an active role in maintaining the hair cell precursor stage.

Table 2. Top upregulated microRNAs in the UB‐OC1 precursor cell line at 33°C (control) versus differentiated cell line at 39°C (High temp).

Figure 10: Inhibition of miR‐210 induces Pou4f3 expression in UB‐OC1 cells.

LNAs against various miRNAs were transfected in UB/OC‐1 cells at 33°C and expression of the hair cell marker Pou4f3 was assessed by rtPCR. rtPCR for GAPDH was used as loading control. Cochlear tissue was used as positive control. Blockade of mir‐210 results in hair cell marker expression.

Overexpression of miR‐210 promotes hair cell formation in organ of Corti explants

Based on the hypothesis that miR‐210 plays an active role in maintaining a progenitor cell type, we speculated if over‐expression of miR‐210 in differentiated cells may reverse their phenotype.

To explore this possibility, we cultured explants of organ of Corti from postnatal day 3 (P3) wild‐ type mice and transduced explants with an adenovirus construct for overexpression of miR‐210

(Adeno5‑mir210). We selected adenovirus vectors since they were previously shown to facilitate transduction of hair cells and supporting cells in mouse inner ear explants and are

45 regarded as a valuable model for developing inner ear gene therapy protocols [77]. To confirm the efficiency of the viral construct, we transduced organ of Corti explants with Adenovirus5 expressing the EGFP reporter gene (Ad5‐EGFP). After 72 hours, only few hair cells were transduced. In contrast, most of the supporting cells expressing Sox2 showed robust EGFP fluorescence following Ad5‐EGFP transduction (Figure 11A). Next, we transduced organ of Corti explants using Ad5‑mir210. Ectopic expression of miR‐210 resulted in formation of additional myosin7A‑positive cells primarily in the outer hair cell area (OHC) (Figure 11B) with a density of

50 cells per 100 µm2 compared to the organ of Corti transduce with Ad5‐EGFP (Figure 11C), thus suggesting that overexpression of miR‐210 facilitates significant formation of new hair cells.

Figure 11: Adenovirus mediated overexpression of miR‐210 in organ of Corti explants.

A. Transduction of adenovirus expressing EGFP in organ of Corti explants shows preferential transduction of cells in the outer hair cell area. B. Transduction of adenovirus expressing miR‐210 in the organ of Corti explants results in new formation of myosin7a positive hair cells preferentially in the outer hair cell region (OHC). C. Increase in myosin 7a‐positive cells per 100 um2 in organ of Corti explants after miR‐210 Adeno5 transduction.

Supporting cells transdifferentiate into hair cells following miR‐210 overexpression

We made use of lineage tracing of the Sox2 supporting epithelial cell marker to study whether the new hair cells formed in the organ of Corti were derived from supporting epithelial cells. For lineage tracing, we crossed Sox2‐CreERT2 knock‐in mice with a mouse line facilitating conditional EGFP expression. Intraperitoneal Tamoxifen injection into the lactating mothers resulted in Cre recombinase‐mediated excision of the floxed‐Stop cassette in double‐transgenic pubs and permanent expression of EGFP in both Sox2 expressing cells and progeny thereof

(Figure 12A). Histological examination of the organ of Corti from P3 double transgenic offspring showed no myosin7a hair cells derived from Sox2 expressing supporting epithelial cells (control,

Figure 12B). In contrast, transduction of Ad5‐mir210 in organ of Corti explants from P3 double transgenic offspring revealed a number of myosin7a positive hair cells derived from Sox2 positive supporting epithelial cells (EGFP and myosin7a double‐positive resulting in yellow fluorescence), thus suggesting that some hair cells had transdifferentiated from supporting cells

(Figure 12C).

Figure 12: Lineage tracing of Sox2 positive supporting cells after miR‐210 overexpression in organ of Corti explants.

A. Schematic representation of mice and breeding used for Sox2 lineage tracing. B. Confocal image showing organ of Corti explants of Sox2CreERT2/R26EGFP double transgenic newborn 4 days after tamoxifen administration. No myosin7a expression is detected in cells from the Sox2 lineage (EGFP, green). C. Organ of Corti explants of Sox2CreERT2/R26EGFP double transgenic mice after Adeno5‐mir210 transduction. A number of cells from the Sox2 lineage express the hair cell marker myosin7a (red).

Putative targets of miR210

We performed an extended computational search to identify potential targets of miR‐210. Using the TargetScan algorithm, we generated a list of 44 conserved predicted transcript targets for the 3’ arms of the mouse and the human miR‐210 sequences. This prediction was extended by adding 6 and 307 miRNA‐mRNA interactions reported in MetaBase for mouse and human miR‐

210‐3p, respectively. Human mRNA interactions were used to annotate the mouse miRNA predictions. These results were combined into a final list of 35 miR‐210 putative targets (Table

3). A similar target prediction analysis was conducted for the miR‐210‐5p arm but did not lead to any significant results. Our 35 predicted miR‐210 targets were compared to the miR‐210 targets previously reported by He et al. [78] and Wang et al. [79] and annotated for their presence or absence in Table 3 (e.g. with “true” or “false”). While some genes in our prediction were previously considered, 18 out of 35 (51, 4%) were newly identified putative targets.

Table 3. List of 35 putative miR‐210 targets obtained from our integrative analysis approach.

Genes were split into five different classes based on the nature of the supporting evidence (literature‐ based or computational only). Literature‐based miRNA‐mRNA interactions retrieved from MetaBase were annotated according to the ‘trust’ of the source into “Present”, “Probably present”, “Not likely present”, or “Absent” (e.g. if not available). Computational inferred interactions identified by TargetScan were classified in two different groups, based on whether targets were predicted for the two species analyzed (“Mouse and Human”) or only one (“Mouse or Human”). Target prediction comparison was done using previously reported prediction analysis of miR‐210 targets in He et al. 2012 and Wang et al. 2014, and was annotated as present=true not present=false. RNA‐Sequencing expression results are reported in the last 2 columns: log2 ratio shows the log 2 expression ratio of the target gene in the UB‐ OC1 precursor cell line at 33° versus differentiated cell line at 39°C log 10 padj shows the log 10 statistical significance of these expression changes after Benjamini‐Hochberg FDR adjustment.

Not in log10 Reported in the Computationally Wang In Wang In He log2 Gene ID Gene symbol padj literature as inferred for et al or et al et al ratio (mouse) (mouse) Class (MetaBase) (TargetScan) He et (2014) (2012) (39 vs 33) (39 vs 33) al

1 72168 Aifm3 Present Mouse and Human TRUE FALSE FALSE ‐0.71 0.05

1 12043 Bcl2 Present Mouse and Human FALSE TRUE FALSE ‐0.5 9.06

1 18033 Nfkb1 Present Mouse and Human TRUE FALSE FALSE ‐0.24 8.16

1 12176 Bnip3 Present Mouse and Human TRUE FALSE FALSE ‐0.72 32.04

1 20423 Shh Present Mouse and Human TRUE FALSE FALSE

1 20852 Stat6 Present Mouse and Human TRUE FALSE FALSE ‐0.39 19.79

1 21416 Tcf7l2 Present Mouse and Human TRUE FALSE FALSE ‐0.06 0.32

2 56336 B4galt5 Probably present Mouse and Human FALSE TRUE FALSE ‐0.02 0.09

2 66383 Iscu Probably present Mouse and Human FALSE TRUE FALSE ‐0.47 15.08

2 12064 Bdnf Probably present Mouse and Human FALSE TRUE FALSE ‐0.99 88.69

2 53417 Hif3a Probably present Mouse and Human FALSE TRUE FALSE 2.25 0.24

2 13638 Efna3 Probably present Mouse and Human TRUE FALSE FALSE 3.09 141.94

2 17992 Ndufa4 Probably present Mouse and Human FALSE TRUE FALSE 0.88 34.15

3 333433 Gpd1l Not likely present Mouse and Human FALSE TRUE TRUE 0.36 10.51

3 381022 Kmt2d Not likely present Mouse and Human FALSE TRUE FALSE #N/A #N/A

3 22661 Zfp148 Not likely present Mouse and Human TRUE FALSE FALSE 0.16 2.84

3 68041 Mid1ip1 Not likely present Mouse and Human FALSE TRUE FALSE ‐0.53 20.07

3 231207 Cpeb2 Not likely present Mouse and Human TRUE FALSE FALSE 1.32 32.14

3 170729 Scrt1 Not likely present Mouse and Human TRUE FALSE FALSE 0.8 0

3 240057 Syngap1 Not likely present Mouse and Human TRUE FALSE TRUE #N/A #N/A

3 74287 Kcmf1 Not likely present Mouse and Human FALSE TRUE FALSE ‐0.64 50.63

3 18013 Neurod2 Not likely present Mouse and Human TRUE FALSE FALSE 0.63 0

4 207393 Elfn2 Absent Mouse and Human FALSE TRUE FALSE 1.98 0.58

4 17258 Mef2a Absent Mouse and Human TRUE FALSE FALSE 0.29 6.21

4 74244 Atg7 Absent Mouse and Human FALSE TRUE FALSE 0.34 9.24

5 69662 2310061I04Rik Absent Mouse or Human FALSE TRUE FALSE ‐0.78 36.81

5 225791 Zadh2 Absent Mouse or Human FALSE TRUE FALSE 0.2 3.44

5 52132 Ccdc97 Absent Mouse or Human FALSE TRUE FALSE ‐0.28 7.63

5 210573 Tmem151b Absent Mouse or Human TRUE FALSE FALSE 2.56 9.58

5 20362 Sept8 Absent Mouse or Human TRUE FALSE FALSE #N/A #N/A

5 433940 Fam222a Absent Mouse or Human TRUE FALSE FALSE #N/A #N/A

5 320717 Pptc7 Absent Mouse or Human FALSE TRUE FALSE ‐0.69 42.05

5 545554 Ankrd34a Absent Mouse or Human TRUE FALSE FALSE 1.2 4.55

5 11515 Adcy9 Absent Mouse or Human TRUE FALSE FALSE ‐0.33 2.88

5 102247 Agpat6 Absent Mouse or Human TRUE FALSE FALSE ‐0.22 6.41

MiR‐trap of miR‐210 associated targets

To physically capture miR‐210 targets, we performed a miRNA pull‐down experiment using the miR‐trap system (Clontech). For this, we co‐transfected UB/OC‐1 cells with a vector for pre‐miR‐

210 expression together with a vector driving the expression of a dominant negative subunit of

RISC that enables miRNA binding to target RNAs but prevents further processing. Following pull‐ down of RISC, captured RNA was isolated and fold‐enrichment of mRNAs was determined by using a qRT‐PCR array for miR‐210 targets (Qiagen) and containing all genes predicted above

(Table 4). To minimize potential artifacts, miR‐210 containing RISC complexes were compared to

RISC complexes pulled‐down from cells transfected with a scrambled sequence. All Ct values were normalized to GAPDH and a 2‐fold enrichment versus control was regarded as a positive result. Of 86 potential miR‐210 targets analyzed, transcripts for 25 genes showed a greater than

2‐fold enrichment (Figure 13) of which 7 were predicted in our computational analysis.

Table 4 : Mouse genes identified in the RISCtrap screens.

The genes confirmed from the prediction list are highlighted in bold letters. RNA‐sequencing expression results are reported in the last 2 columns: log2 ratio shows the log 2 expression ratio of the target gene in the UB‐OC1 precursor cell line at 33°C (control) versus differentiated cell line at 39°C (High temp); log 10 padj shows the log 10 statistical significance of these expression changes after Benjamini‐Hochberg FDR adjustment.

log10 padj log2 ratio (39 vs 33) (39 vs 33) gene_id current_symbol description 11479 Acvr1b activin A receptor, type 1B 0.71 21.69 12064 Bdnf brain derived neurotrophic factor ‐0.99 88.69 108699 Chn1 chimerin 1 ‐0.2 0.13 211922 Dennd6a DENN/MADD domain containing 6A #N/A #N/A 14357 Dtx1 deltex 1 homolog (Drosophila) 0.72 0.09 family with sequence similarity 222, 433940 Fam222a member A #N/A #N/A 14782 Gsr glutathione reductase ‐0.45 20.85 15245 Hhip Hedgehog-interacting protein ‐1.66 4.93 15394 Hoxa1 homeobox A1 ‐0.5 0.86 IscU iron-sulfur cluster scaffold 66383 Iscu homolog (E. coli) ‐0.47 15.08 74287 Kcmf1 potassium channel modulatory factor 1 ‐0.64 50.63 potassium channel tetramerisation domain 216858 Kctd11 containing 11 1.15 78.04 NADH dehydrogenase (ubiquinone) 1 17992 Ndufa4 alpha subcomplex, 4 0.88 34.15 PTC7 protein phosphatase homolog (S. 320717 Pptc7 cerevisiae) ‐0.69 42.05 repulsive guidance molecule family 244058 Rgma member A ‐0.25 5.57 104001 Rtn1 reticulon 1 1.36 0.12 SET domain containing (lysine 67956 Setd8 methyltransferase) 8 ‐0.44 25.81 solute carrier family 25 (mitochondrial 67582 Slc25a26 carrier, phosphate carrier), member 26 0.02 0.03 synaptic Ras GTPase activating protein 240057 Syngap1 1 homolog (rat) #N/A #N/A transcription factor 7 like 2, T cell specific, 21416 Tcf7l2 HMG box ‐0.06 0.32 210573 Tmem151b transmembrane protein 151B 2.56 9.58 56338 Txnip thioredoxin interacting protein ‐0.43 14.03 213742 Xist inactive X specific transcripts 1.31 0.88 21769 Zfand3 zinc finger, AN1-type domain 3 ‐0.15 2.98

Figure 13. Pull‐down of miR‐210 target RNAs.

Target RNAs for miR‐210 were pulled down by MirTrap and candidate targets were identified by qPCR array. A minimum of 2‐fold enrichment was used as a positive identification of a microRNA target. Y axes represent relative fold enrichment.

4.1.5 Discussion

Sensorineural hearing loss is the most common sensory deficit in the world and as the population continues to age and expand, the number of patients who suffer from serious hearing loss is consistently increasing. Unlike birds, reptile and fish, mammals are unable to regenerate auditory hair cells after damage. To understand the failure of mammalian hair cell regeneration, there is a need to understand the molecular mechanisms that underlie cell division control and hair cell differentiation during embryogenesis and in postnatal mouse. A number of studies have shown that miRNAs are involved in the development of inner ear hair cells [42]. These studies have led to the idea that miRNAs may offer a novel therapeutic approach, since they have been implicated in promoting regeneration by regulating genes that control differentiation of hair cells.

This study was designed to profile changes in the expression of miRNAs in the inner ear cell line

UB/OC1 as it is induced to differentiate. UB/OC1 is a conditionally immortal cell line selected from the cochlea of an Immortomouse at embryonic day 13 (E13), to represent the developmental stage at which hair cell progenitors cease to proliferate and start to differentiate

[49]. The hair cell precursors permit functional studies of cochlear genes and allow the exploration of key factors and signaling pathways that may facilitate therapeutic stimulation of hair cell replacement.

RNA sequence analysis of the UB/OC1 cell line led to the identification of 99 differentially expressed miRNAs. Eight of the most differentially expressed microRNAs, when compared between the precursor and differentiated UB/OC1 cells, were selected for further validation. To

55 verify their function, the selected microRNAs were inhibited by transfection of LNA antagonists into the precursor cell line and the effect they produced was characterized. The characterization monitored if this inhibition could lead to forced differentiation of the cell line without inactivation of the T‐antigen in the cells. The functional study identified miR‐210 inducing differentiation of the UB/OC1 to a hair cell like state, as shown by the expression of Pou4f3, which is the first hair cell marker to be expressed during normal formation of the inner ear hair cell [80, 81]. To further characterize this observation, miR‐210 was overexpressed using an

Adenovirus expression system in organ of Corti explants. Organ of Corti explant systems are an excellent model regarding gene expression, stereocilia motility, cell biology of the inner ear. In particular, organ of Corti explants are ideal models to explore approaches for inner ear hair cell regeneration and developing inner ear gene therapy [77].

Adeno5 mediated overexpression of miR‐210 in the cochlear explants showed a higher number of ectopic hair cells in the great epithelial ridge of the organ of Corti. Lineage tracing was used to discover the origin of the newly formed hair cells. The results show that a number of myosin7a positive hair cells were also found to be positive for Sox2 in this explant model.

Consistent with this observation is the hypothesis that the newly formed hair cells have been formed by trans‐differentiation of supporting cells.

MiR‐210 has been linked to several pathways and regulation of hypoxia regulated genes [82], mitochondrial metabolism, DNA repair and apoptosis [81]. MiR‐210 has also been reported to be involved in the cell cycle regulation via activating the myc pathway [83]. It is also overexpressed in cells affected by cardiac disease and tumors [81]. The wide multifunction of miR‐210 makes it a novel potential tool for gene therapy in many diseases. For example, a novel

56 study shows that lentiviral mediated overexpression of miR‐210 induce angiogenesis and neurogenesis which is associated with VEGF upregulation in adult brain and could be used as treatment of cerebral ischemia [84].

In order to devise a list of candidate transcript targets for miR‐210 that could play a role in the hair cell differentiation, we performed a target prediction analysis with TargetScan and reviewed previously published miR‐210 interactions retrieved from Metabase. In addition, miR‐

210 predicted targets were compared to previous target prediction analyses done by He et al.

2012 [78] and Wang et al. 2014 [79]. In He et al., a combination of genome‐wide scale analysis and bioinformatics was used to identify functionally related target genes of miR210, which were found to be down‐regulated in CNE cells under hypoxia condition. Using this approach, they identified two functional gene groups, of which one was involved in cell cycle regulation and the other related to RNA processing. Wang et al. used four algorithms (Target scan, PicTar, miRDB and miRanda35–38) to predict miR‐210 targets and combined their targets with miR‐210 targets previously reported in the literature. We used a similar prediction approach like in Wang et al.

2014, and a number of candidates of our analysis is matching also with the Wang target prediction analysis.

Potential miR‐210 direct targets were assessed for direct association with miR‐210 within RISC complexes. We immunoprecipitated RISC complexes enriched for miR‐210 targets and quantified the enrichment of a set of transcripts including transcripts from the candidate list. A number of targets validated in the RISC immunoprecipitation assay were associated with the generation of neurons, neurogenesis and nervous system development (13 of 24 gene targets).

Some targets were already described in the inner ear, like the brain‐derived neurotrophic factor

(BDNF) and homeobox A1 (Hoxa1). BDNF is required for the development and maintenance of normal innervation of hair cells. It is expressed in hair cells and supporting cells and is known to be involved in the repair of the auditory nerve following damage or nerve cell loss [85] [86].

The transcription factor Hoxa1 has also been reported to be expressed in the developmental otic epithelium and has a strong contribution to the development of the inner ear [87, 88].

Microarray analysis of Hoxa1 null embryo mice revealed downstream targets of Hoxa1 necessary for early inner ear development, for example fibroblast growth factor receptor‐3

(Fgfr3) [88]. Fgfr3 is known to play a role in the in the differentiation of hair cells and supporting cells, and loss of FGFR3 leads to excess hair cell development in the mouse organ of Corti [89].

Interestingly, Fgfr3 is expressed in the developing progenitor cells and mature supporting cells of the organ of Corti. Inhibition of Fgfr3 in the organ of Corti increases the number of hair cells.

This increase was not associated with increased proliferation, suggesting that the inhibition of

Fgfr3 leads to the direct transdifferentiation of supporting cells into hair cells [90].

Another interesting target found to be associated with miR‐210 is Kctd11 (REN). Kctd11 is a regulator of neuronal differentiation. It was shown that Kctd11 enhances p27Kip expression in neuronal progenitors, causing growth arrest and further neural differentiation [91]. This mechanism of terminal differentiation in neurons is similar to hair cell differentiation in the organ of Corti. P27Kip1 induces cell cycle arrest of hair cells progenitors in the organ of Corti and causes terminal differentiation. It was also shown that Kctd11 inhibits sonic hedgehog (Shh) dependent events, thus antagonizing Shh‐induces effect on the proliferation and differentiation of granule cell progenitors [92]. Sonic hedgehog signaling is essential for auditory cell fate determination. Shh accelerates inner ear progenitor cell proliferation and hair cell formation

[92]. Our hypothesis is that with the inhibition Kctd11 by miR210, hedgehog antagonism is impaired and promotes in this way hair cell formation.

Deltex 1 (DTX1), another validated candidate found in our assay, is an active regulator of the

Notch signaling pathway [93]. DTX1 interacts with the intracellular domain of Notch1 and mimicks the action of DA‐Notch1 to inhibit the transcription factors involved in cell differentiation of neuronal progenitors. So, a possible downregulation of DTX1 could inhibit the

Notch pathways downstream and promote transdifferentiation of the supporting cells into hair cells.

An understanding of the genetic elements involved in the differentiation of inner ear hair cells could be essential for developing therapeutic approaches to restore hearing loss. We have shown that the combination of an inner ear cell line derived from a precursor cell stage and next generation sequencing technology can yield robust insights into such programs. The data allowed us to identify expression of miR210, a microRNA involved in different cell functions, as a promising factor of new hair cell formation via trans‐differentiation of supporting cells in the organ of Corti. Expression of miR‐210 therefore offers a promising opportunity for therapeutics in the context of hearing loss.

5. Identification and characterization of Atoh1 modulators

5.1 MiR‐152 can epigenetically modulate and reactivate the transcription factor Atoh1

5.1.1 Abstract

Disabling hearing loss affects over 5% of the world’s population, resulting in reduced ability to communicate, leading to loneliness, isolation and frustration. Impaired hearing can be a consequence of advancing age, acoustic trauma or side effects from drug treatments (cytotoxic drugs or antibiotics). Such insults result in irreversible damage to the sensitive hair cells inside the inner ear, for which there is currently no regenerative treatment to restore these hair cells.

However, it is possible to force regeneration of damaged hair cells by over‐expression of the transcription factor Atoh1 in the inner ear, resulting in trans‐differentiation of supporting epithelial cells into sensory hair cells.

The aim of this project was to identify genetic elements that, when modulated, lead to reactivation of Atoh1 expression. Using a human intestinal epithelial cell line (SW480), engineered with a luciferase reporter gene under the control of the Atoh1 regulatory locus, we screened a library of miRNA expressing plasmids. This led to the identification of miR‐152 acting as a factor leading to reactivation of Atoh1 gene expression. Reactivation and expression of

Atoh1 was associated with reduced DNA methylation of the enhancer region in the Atoh1 locus.

These results are consistent with the suggestion that miR‐152 acts by inhibiting expression of

DNA‐methyltransferases (DNMTs). To further characterize the link between miR‐152 and Atoh1 activation in the inner ear organ of Corti, miR‐152 was overexpressed in cochlear explants using a viral vector. No activation in the Atoh1 supporting cells was seen in the organ of Corti explants

60 after overexpression of miR152. These results provide evidence that miR‐152 activates Atoh1 via epigenetic modulation in the intestinal colon cancer cell line but shows no effect in the reactivation of Atoh1 in the organ of Corti. Further investigations are needed to study whether miR‐152 might play a role in goblet differentiation or as a tumor suppressor in colon cancer cells.

5.1.2 Introduction

Hearing loss is the most common sensory defect in developed countries, for example, more than 5% of people in the US are suffering from hearing problems, with the prevalence growing

[94]. Hearing loss is a disorder caused many different factors with many genetic and environmental factors contributing to hearing loss [95]. Hearing loss is also age related, as about

4% of people less than 45 years of age suffer debilitating hearing loss, while about 34% of those over 65 years suffer from such hearing loss [94].

In many cases, hearing loss is related to the degeneration and death of inner ear hair cells and their associated spiral ganglion neurons, e.g. [96]. Auditory hair cells of the inner ear are vulnerable to damage from acoustic over‐stimulation, premature birth, infection, ototoxic drugs and aging [97]. Non‐mammalian vertebrates can replace hair cells through transdifferentiation of epithelial supporting cells; such spontaneous regenerative capacity has been lost in mammals

[55]. In mammals, supporting cells can be forced to transdifferentiate into new hair cells by the forced over‐expression of the basic helix‐loop‐helix (bHLH) transcription factor Atoh1 [22].

Atoh1 has been shown to have diverse and sometimes opposite roles; it was first described as having a role in neurogenesis but has also been described as an oncogene and a tumor suppressor, depending on the context, reviewed by Mulvaney and Dabdoub [20] . The

Drosophila homologue atonal (ato) was first shown to be involved in the development of the chordotonal organs which include the sensory bristles, hairs and papillae. Subsequently, the mouse Atoh1 gene was cloned from a screen for ato homologues [98]. The timing and location of Atoh1 expression was monitored in mice where it was found to have a role in the development of the sensory cells of the organ of Corti in the inner ear [99], however, after the

P10 stage in mouse development expression of Atoh1 is repressed, possibly by epigenetic regulation [100].

The diverse roles of Atoh1 come about because of its ability to interact with a number of different transcription factors [101]. Indeed, Atoh1 interacts with a number of different signaling pathways including sonic hedgehog, Notch, TGF beta, Wnt as well as JNK/ MAPK signaling [101]. In intestinal cells, Atoh1 has a role in the differentiation and maintenance of secretory cells (such as goblet, Paneth and enteroendocrine cells). Atoh1 null mice lack such cells and the intestinal epithelial cells are trapped in a proliferative state [102]. One of the mechanisms by which Atoh1 prevents proliferation in gut cells is by interacting with DNA methyltransferases to induce goblet cell differentiation and also by positively regulating apoptosis through the JNK/MAPK pathway [103].

Recently, Atoh1 directed maintenance of hair cell fate in the organ of Corti has been reported for the miR‐183 family of micro‐RNAS (miRNAs)[42]. MiRNAs are endogenous, small (20–23 nt) non‐coding RNAs that bind to complementary sequences within target messenger RNA (mRNA)

62 transcripts and typically result in translational repression or degradation of the target mRNA and gene silencing [104]. MiRNAs are collectively predicted to target ~60% of all genes and each miRNA is expected to repress hundreds of target genes, suggesting that they may serve as tools for pharmacological intervention [105]. miRNAs are a vital part of genetic regulation and exhibit a wide range of biological functions including cell differentiation, proliferation, apoptosis, metabolism, and self‐renewal. Recent studies have established a direct correlation between miRNAs and the coordinate regulation of the expression of mRNAs involved in common pathways.

The aim of this study was to identify genetic elements that when modulated lead to reactivation of Atoh1 gene expression. As part of these efforts, a small set of plasmids expressing micro‐

RNAs were tested. For this purpose, a reporter cell line was engineered with a luciferase reporter gene under the control of the entire Atoh1 regulatory locus, so that transcriptional activation leads to increased luciferase expression. One of the actives identified from this screening effort was further characterized for its molecular pathway in vitro and in ex‐vivo. Our results provide evidence that miR152 activates Atoh1 via epigenetic modulation in the intestinal colon cancer cell line but shows no effect in the reactivation of Atoh1 in the organ of Corti.

5.1.3 Materials and methods

Reporter human cell line generation

BAC transgene:

For the generation of the reporter cell line, the human BAC RP11‐586I3 was used. The BAC contains 145 kb upstream sequences including the Atoh1 promoter as well as 21 kb of the 3’ region. Firefly luciferase cDNA and the PGK‐neomycin cassette was integrated into the Atoh1

ATG.

Generation of stable cell lines:

For the generation of stable expression clones, the human intestinal epithelial cell lines SW480

(ATCC CL‐227) was transfected with the linearized BAC reporter (described before) using the

Lipofectamine2000 reagent (Life Technologies). Individual cell clones were selected with 500

μg/ml G418 (Life Technologies). Clones were selected based on the activation of Atoh1 luciferase after treatment with 15 nM of the pan‐HDAC inhibitor LBH589, 100uM

DNAmethyltransferase inhibitor 5‐Azacytidine (Sigma), 10 uM γ‐secretase inhibitor Compound E

(Calbiochem) and 10 uM GSK‐3α/β inhibitor BIO (Sigma).

MicroRNA library generation

The genomic region of the miRBase‐documented pre‐miRNA and 400 bp up/downstream was cloned as intron into a GFP‐expression vector with beta‐globin splice sites. In cases where an exon fell within the +/‐ 400bp flanking region, the additional genomic DNA was truncated such

64 that there was a distance of at least ~20 bp to the corresponding boundary to avoid the inadvertant introduction of additional splice sites. Similarly, in the case of polycistronic expression of several miRNAs, the insert was trimmed such that the total length was below 1kb, if possible. The beta‐globin intron was inserted in the standard splicing recognition sites of the

GFP coding sequence (Figure 14). Thus, each time when the intron is made and spliced correctly, the cells express GFP, a proof that a piece of RNA was synthesized and processed (the intron) in the cells. The cloning site in the intron was made by site directed mutagenesis . A total number of 32 microRNAs expressing plasmids were cloned (Table 5).

Figure 14: Schematic representation of one of the constructs used in the microRNA library

Table 5: List of the microRNA plasmids cloned for the library screening

Sample no microRNA name

1 let‐7c

2 miR‐1‐1

3 miR‐7

4 miR‐9‐1

5 miR‐15a_miR‐16‐1

6 miR‐17‐5p_miR‐18a miR‐19a_miR‐19b_miR‐20a_miR‐92

7 miR‐21

8 miR‐24‐1_miR‐189

9 miR‐29a

10 miR‐30a

11 miR‐93

12 miR‐96

13 miR‐101

14 miR‐103

15 miR‐106a

16 miR‐107

17 miR‐122a

18 miR‐133a‐1

19 miR‐139

20 miR‐143

21 miR‐144

22 miR‐146a

23 miR‐146b

24 miR‐152

25 miR‐153

26 miR‐155

27 miR‐181a‐1

28 miR‐185

29 miR‐195

30 miR‐204

31 miR‐206

32 miR‐211

33 miR‐302a_miR‐302b_miR‐302c_miR‐302d_miR367

34 miR‐324

35 miR338

36 miR‐373_miR‐372_miR‐371

37 miR‐377

38 miR‐449_miR‐449b

39 miR‐542

Screening assay

SW480‐Atoh1Luc cells were sub‐cultured in the media (RPMI 1640, 10% FBS, 500 ug/ml G418,

Life technologies ) and plated at a density of 4x103 cells per well in 384‐well white matrix plates

(Thermo Scientific) and returned to the incubator overnight at 37°C, 5% CO2. The next day, the

67 cells were transfected with microRNA plasmids library with the transfection reagent Genjet plus

(Signagen) and returned to the incubator at 37°C, 5% CO2. After 72 hours of stimulation, Steady glo (Promega) was added to the plates and incubated for 5 mins at room temperature and read on the Envision plate reader, 0.1 sec/well. The data were captured in Excel and analyzed in

Graph Pad prism.

Generation of a stable cell line overexpressing miR‐152

For the generation of stable expression clones, the human intestinal epithelial cell line SW480

(ATCC CL‐227) was transfected with the same plasmid used in the microRNA screening using the

Lipofectamine2000 reagent (Life Technologies). Individual cell clones were selected based on the EGFP and miR‐152 expression.

Mass spectrometry analyses

SW480 cell pellets were lysed in 200 uL RIPA buffer (Cell Signaling Technology) containing protease inhibitors (Pierce). Protein concentrations were determined using the BCA kit (Pierce).

Proteins (400 ug) were reduced with 5 mM dithiothreitol for 15 min at 60°C and alkylated with

10 mM iodoacetamide for 1 hour at 25°C in the dark. Trypsin (MS grade, Promega) was added at a 1:25 ratio (trypsin:protein) by mass and digestion was carried out overnight at 37°C. The peptide mixtures were loaded onto C18 cartridges (Empore C18, 3M), previously conditioned using methanol, followed by 70% acetonitrile/0.1% triflouroacetic acid and 0.1% triflouroacetic acid. The cartridges were washed using 0.1% triflouroacetic acid and the peptides were eluted in 70% acetonitrile and evaporated to dryness.

Crude stable isotope‐labeled peptides were synthesized with 13C615N2‐lysine or 13C615N4‐ arginine residues by Thermo Fisher Scientific (Rockford). Desalted digests were reconstituted in a mixture containing internal peptide standards and were injected onto an Agilent 1200 Series

HPLC for offline fractionation (Agilent Technologies). Mobile phases consisted of 20 mM ammonium formate pH10 (Solvent A) and acetonitrile pH10 (Solvent B). Separation was achieved using an Xterra C18 (3.5 um, 2.1 x 150 mm) reversed‐phase column (Waters

Corporation) heated to 40°C. A gradient of 2‐40% Solvent B over 45 minutes at 200 uL/min, followed by column washing for 4 minutes at 90% Solvent B and equilibrating for 10 minutes at

2% Solvent B was used. Three wavelengths (215 nm, 254 nm, and 280 nm) were monitored.

MRM experiments were performed on a TSQ Vantage triple‐quadrupole instrument (Thermo

Scientific) interfaced with an Eksigent NanoLC ultra 2D plus and NanoLC‐AS1 (Eksigent

Technologies). Tryptic digests were separated on a PicoFrit column (75‐um i.d., 15‐um tip opening; New Objective) packed in‐house with 12 cm of Magic C18 3‐um reversed‐phase resin

(Michrom Bioresources). Mobile phases consisted of 2% acetonitrile in 0.1% formic acid (Solvent

A) and acetonitrile in 0.1% formic acid (Solvent B). Peptides were separated using a gradient of

2‐40% Solvent B over 25 minutes at 250 nL/min, followed by column washing for 3 minutes at

80% Solvent B and equilibrating for 10 minutes at 2% Solvent B. The mass spectrometer was operated in positive ion mode with spray voltage of 1.500 V, capillary temperature of 225°C, Q1 and Q3 resolution settings of 0.70 FWHM, and a cycle time of 1 sec. Collision energy (CE) parameters were calculated using linear equations in Skyline1 (MacCoss Lab Software) and collision cell gas pressure of 1 mTorr was used for fragmentation.

MRM transitions for monitoring were selected by choosing the three most intense transitions from doubly or triply charged precursor ions and singly charged fragment ions of the y‐ and b‐ ion series. Using peptide retention times, scheduled methods were set up with 6 min detection windows. Data were processed and methods established using Skyline software2 (MacCoss Lab

Software). MRM data was manually inspected to ensure correct peak detection, accurate integration, consistency of transition ratios, as well as the absence of interferences. Ratios reported by Skyline include the sum of peak areas from the three measured MRM transitions from the endogenous sample divided by the response from the internal standard. Protein fold change was calculated by comparing expression in the SW480 cells overexpressing mir152 with the control cell line.

Detection of DNA methylation

Genomic DNA was extracted from cells with the DNeasy Blood&Tissue kit (Qiagen) and subjected to sodium bisulfite modification by using the EZ DNA Methylation Gold Kit

(Zymoresearch) EpiTect bisulfite kit. The amplication of the bisulfite converted enhancer region was carried out with GoTaq hot star polymerase system (Promega, WI, USA), using the following primers:

Forward 5’ATTAGAGAGYGGTTGATAATAGAGGGGTTG3’

Reverse 5’CTACRCAACTAAAAACCAAAAAAACCC3’

Product size: 291 bp, CpGs in product: 25

PCR products were purified by gel electrophoresis and subcloned into the pcR2.1 TOPO TA

Vector System (Invitrogen). The amplicons were sequenced by Sanger sequencing at the

Novartis Institutes for BioMedical Research. For the methylation analysis, the software BISMA

(http://biochem.jacobs‐university.de/BDPC/BISMA) was used.

Adenovirus generation

The adenoviral vector was constructed using the Virapower Adenoviral Expression system (Life

Technologies). Briefly, the pre‐miR152 was inserted using the Gateway system (Life Technogies) into an adenoviral backbone plasmid (pAd/CMV/V5‐DEST). The linearized Adeno plasmid was transfected into the adenovirus packaging 293A cell line (Life Technologies), following the

Viropower system protocol. Viruses were extracted from infected cells 10 days after infection by three freeze–thaw cycles and stored at − 80 °C unl needed. Viral ters were measured by standard end‐point dilution assay using the 293A cells. Furthermore, mir152 expression was confirmed by measuring the RNA expression level of the infected 293A cells (Lifetechnologies) by Taqman analysis (Figure 15).

Figure 15: Fold increase of miR‐152‐3p expression after CMV‐miR152 adenoviral transduction of HEK293A cells.

Reporter mice generation

Atoh1‐EGFP‐IRES‐CreERT2 mice were generated by homologous recombination in C57BI/6 ES cells, targeting EGFP‐IRES‐CreERT2 cassettes to the ATG of Atoh1. Successful targeting was confirmed by Southern blot and PCR analyses. Targeted C57Bl/6 ES cells were injected into

BALB/c host blastocysts, which were then transferred into pseudopregnant B6CF1 foster mothers. Chimeric mice were mated with C57Bl/6 wt mice, and germline transmission of the targeted ES cells was confirmed using genotyping of EGFP. For the study, timed matings were performed using the heterozygous Atoh1‐EGFP crossed to C57Bl/6 mice. Protocols, handling and care of the mice conformed to the Swiss Federal law for animal protection

Cochlear explants assay

Inner ear organ cultures were established at postnatal day 3 from the Atoh1 –EGP reporter mice. After removal of the cochlear bone and dissection out of the lateral wall in Hank's balanced salt solution supplemented with 10 mm Hepes buffer (pH 7.3) on ice, the sensory epithelia were transferred onto a 0.4 cm well culture plate (BD Biosciences Discovery), precoated with celltak (BD) containing Dulbecco's modified Eagle's medium–F12 (Gibco), supplemented with 1 μg/mL ampicillin (Gibco) and 1% FBS (Gibco) and then kept at 37 °C in 5%

CO2 for 1 day. The next day, the cochlear cultures were transduced with the Adeno5 virus miR‐

152 or treated with the compound pan‐HDAC inhibitor 50 nM LBH589 (Novartis), 100 uM 5‐

Azacytidine (Sigma). 1x109 virus particles were diluted in the inner ear explants media and added to the explants. The inner ear organs were kept in culture at 37°C in 5% CO2 for additional 3 days before the analysis.

Cochlear explants whole mount immunostaining

Cochlear explants were fixed with 4% paraformaldehyde at room temperature (22‐24°C) for 30 min. The specimens were then blocked with the blocking buffer containing 5% of donkey serum

(Sigma) and 0.1% of Triton X‐100 in phosphate‐buffered saline (PBS) at room temperature. After

1 hour incubation, the blocking buffer was removed and antibodies to myosin 7a (1:500;

Biosciences), EGFP (1:500;Invitrogen) or Sox2 (1:100; Santa Cruz), diluted in the blocking buffer, were added to the explants and incubated overnight at 4°C. The next day the specimens were incubated for one hour at room temperature with the secondary antibody (Invitrogen), diluted in the blocking buffer at 1:500. After several rinses in PBS, the slides were finally mounted on a slide with Fluorsave (Calbiochem).

Cell count analysis

The cell populations were assessed on a Zeiss Z.1 confocal microscope (Zeiss, München), using a

40× oil immersion objective. The hair cells were counted in 100 μm segments along the length of the cochlea. Each group had at least 3 different cochlear explants and each explant was sampled in 5 areas. All cell counts were performed by manually analyzing the confocal images.

The values are expressed as the mean ± standard error and using a student paired t‐test for statistical analysis, a P<0.05 was considered to indicate a statistically significant difference.

5.1.4 Results

The reporter cell line SW480‐Atoh1 BAC Luciferase responds to epigenetic modulation

The aim of this project was to identify reagents that might lead to the expression of Atoh1 in the organ of Corti and hopefully the trans‐differentiation of support cells into new sensory hair cells.

In the absence of a suitable inner ear cell line (in which Atoh1 expression would be repressed), the SW480 colon cancer cell line was selected, based on its repression of endogenous Atoh1 expression. An Atoh1‐luciferase bacterial artificial chromosome (BAC) reporter construct was generated in which both the 5’ and 3’ regulatory regions of Atoh1 were present (Figure 16A).

The suitability of these cells as a model for screening for modulators of Atoh1 expression was then tested by monitoring the expression of both the reporter gene and the endogenous Atoh1 in the presence of GSK‐3α/β‐, γ‐secretase‐, or HDAC inhibitors (Figure 16B) which had previously been show to regulate Atoh1 expression in colon cancer cell lines. As a further control, the response of the luciferase reporter was also monitored by its reactivation by 5‐azacytidine which is known to inhibit DNA methyltransferases (Figure 16C). The reactivation of the reporter gene by these treatments is consistent with the reporter being epigenetically silenced.

Figure 16: Response of the Atoh1 endogenous and reporter gene to epigenetic repression

A Schematic representation of the Atoh1 BAC construct inserted in the SW480 cell line.

B Left: Activation of the Atoh1 luciferase reporter cell line after treatment with BIO, CompE and pan HDACi LBH589. Right: LBH589 shows to activate the endogenous Atoh1 in the SW480 reporter cell line. LBH589 was used as positive control in the assay.

C Activation of Atoh1 in the reporter cell line after the treatment with DNA methyltransferase inhibitor 5‐Azacytidine .

MiR‐152 and miR‐153 reactivate Atoh1 in the SW480‐Atoh1 Luc cell line

In order to identify Atoh1 modulators, a library of 32 different microRNA expressing plasmids

(Table 5) were transfected into the SW480‐Atoh1 BAC luciferase reporter cell line. Luciferase expression was measured after 72 hours induction. The screening was performed in duplicates.

Two microRNA expressing plasmids, miR‐152 and miR153, were found to activate the Atoh1 reporter gene in this cell line (Figure 17).

200 miR-152 150 miR-153

100

-50 50 100 150 200 -50 Techreplicate1 (Normalized activity %)

Figure 17: Modulation of Atoh1 reporter gene expression by miRNAs.

The screen identified miR‐152 and miR‐153 to significantly and reproducibly activate the Atoh1 reporter in this cell line.

Overexpression of miR‐152 reduces DNA methyltransferases in SW480 cells

The miR‐152 was recently described to regulate the expression of DNA methyltransferases

(DNMT’s) [106]. In order to evaluate a possible correlation of miR‐152 with DNMT’s, and because micro‐RNAs are reported to work by modulating the efficiency of mRNA translation

(not the levels of mRNA present in the cell), protein levels were monitored with or without overexpressing mir‐152 to determine if the DNA methyltransferases were in fact modulated.

Because of the poor transfection efficiency of the recombinant SW480 cell line (and the fact that the cell line was apparently unstable after passage 25, with the cells growing more slowly and adapting a different morphology), it was decided to make a stable cell line expressing miR‐

152 in the parental SW480 cell line. This was accomplished using transfection of the miR‐152 expression construct. Two expression clones were grown up and used for a mass spectrometry

(MS) based assay to monitor changes in the DNMT protein levels compared to the parental cell line. It was necessary to use the MS based assay to monitor DNMT levels because the commercially available antibodies were either not able to detect DNMT in these cells or the number of background contaminating bands made it difficult to clearly identify and measure the levels of DNMT proteins present in the samples. The results presented in Figure 18 show that

DNMT’s are downregulated at the protein level by overexpression of miR‐152 in two different clones, as compared to the parental cell line. This is consistent with the report of Qing et al.

[106], which suggests that miR‐148a /152 is coordinately regulating DNMT’s expression.

Figure 18: Mass spectrometry results showing DNMT protein levels with and without overexpressing miR‐152 in two different clones. Expression of DNMT decreases after overexpressing mir‐152 in the SW480 cells.

Overexpression of miR‐152 changes the methylation pattern of the Atoh1 enhancer

The initial characterization of the reporter cell line suggested that the Atoh1 enhancer is regulated at the level of DNA methylation (Figure 19). The enhancer region in the Atoh1 locus is a conserved sequence and its activation plays a crucial role in the transcription of Atoh1 in the development of the sensory hair cells. To evaluate if miR‐152 overexpression has an effect on the epigenetically silenced Atoh1 enhancer, bisulfide sequencing and analyses of the DNA methylation pattern was conducted. The results are consistent with miR‐152 overexpression changing the methylation profile at the Atoh1 enhancer (Figure

19).

Figure 19: Regulation of Atoh1 by DNA methylation.

Bisulfite sequencing of the Atoh1 locus shows that the enhancer is hypermethylated.

Overexpression of miR‐152 in the SW480 modulates the methylation pattern of the Atoh1 enhancer.

Overexpression of miR‐152 shows no reactivation of Atoh1 silenced cells of the organ of Corti

To further analyze the link between miR‐152 and the reactivation of Atoh1 in the inner ear, we cultured explants of organ of Corti from postnatal day 3 (P3) Atoh1‐EGFP reporter mice and transduced them with an Adeno5 viral construct overexpressing the miR‐152. Atoh1 is expressed in the hair cells until postnatal day 10. Overexpression of miR‐152 resulted in no

79 ulterior Atoh1 and Myosin7a positive hair cells. The organ of Corti explants were also treated with the DNA methyltransferase inhibitor 5‐Azacytidine and no ulterior positive Atoh1 hair cells were observed compared to DMSO. Only the pan HDAC inhibitor LBH589 induced significant new hair cell formation in the organ of Corti explants (Figure 20), suggesting that the modulation of histones might play an important role in the reactivation of Atoh1 in the inner ear.

Figure 20: Adenovirus mediated overexpression of miR‐152 in organ of Corti explants.

No increase in Atoh1‐positive cell number per 100 um2 in the organ of Corti explants was found after miR‐152 Adeno5.

5.1.5 Discussion

Atonal homolog 1 (Atoh1) is a transcription factor that plays an important role in the differentiation and regeneration of sensory hair cells [99]. This work describes the identification of miR‐152 as a regulator of the expression of Atoh1. This was found using a luciferase reporter gene construct stably integrated in the colon cancer cell line SW480. The overexpression of miR‐152 reduced the protein levels of DNMT1 and other DNA methyltransferases. Due to difficulties in the quantification of protein levels downregulated by microRNAs, it was necessary to develop a mass spectroscopy method that directly measures the DNA methyltransferase protein levels. This experiment was conducted using transfection of a vector expressing miR‐152 under control of the CMV promoter into the parental cell line from which two miR152 clones were used in this experiment. MiR‐152 was shown to significantly reduce the levels of DNA methyltransferases, apparently in a coordinated manner.

This observation is consistent with previous reports that in ovarian cancer and in NiS transformed cells, miR‐152 is capable of downregulating the expression of DNMT1 [107]. DNA methylations give rise to transcriptional silencing of genes, in particular for tumor suppressor genes. Like for Atoh1 in the colon cancer cell line, this may play an important role during carcinogenesis. DNA methylation mainly occurs at the C5 position of CpG dinucleotides by three

DNA methyltransferases (DNMTs), namely DNMT1, DNMT3A, and DNMT3B. DNMT3A and 3B are responsible for de novo DNA methylation, while DNMT1 is the maintenance methyltransferase. The observation that the other DNMTs are also being coordinately regulated

81 may indicate a role in coordinating the regulation of DNA methylation in cells. In this regard, it is of note that expression of miR‐152 may create a feedback loop as DNA methylation of the miR152 promoter controls its expression [108] .

Subsequently, it was possible to detect that in the SW480 cell line, miR152 expression leads to reduction of DNA methyltransferase protein levels, which in turn seems to lead to a reduction in the levels of DNA methylation of the Atoh1 enhancer. This is consistent with previous reports that have described Atoh1 expression being repressed by methylation of Atoh1 in colorectal cancer [109]. These data suggest that methods to increase miR‐152 expression in colorectal cancers may serve as means to target multiple DNA methyltransferases for the treatment of such cancers.

Atoh1 has been shown to have a pivotal role in the development of the inner ear and during mouse development becomes repressed after the P10 stage of development. This repression may be a consequence of silencing factors such as Hes or possibly of methylation of the Atoh1 enhancer in an analogous fashion to repression of Atoh1 in colorectal cancers [109]. In addition, forced expression of Atoh1 in the inner ear has been shown to lead to regeneration of hair cells and to improve hearing [22]. Therefore, the effect of expression of miR‐152 (possibly leading to increased Atoh1 expression) on the inner ear was tested in organ of Corti explants using a viral expression vector. The hypothesis was that miR152 leads to a downregulation of DNA methyltransferases and a subsequent increased expression of Atoh1 that would then lead to a transdifferentiation of the supporting cells into hair cells. No additional hair cells were found after overexpression of miR‐152 in the organ of Corti, suggesting that miR‐152 may play a role in the regulation of Atoh1 in the colon cancer cell. Atoh1 is known to act as tumor suppressor gene

82 which antagonizes tumor formation and growth by regulating proliferation and apoptosis [110].

Atoh1 is silenced in 80% of human colorectal cancers by both CpG island methylation and genetic microdeletion. Furthermore, colorectal cancer patients show genetic and epigenetic

Atoh1 loss of function mutations [110].

These results provide evidence that miR152 activates Atoh1 via epigenetic modulation in the intestinal colon cancer cell line but shows no effect in the reactivation of Atoh1 in the organ of

Corti. Further investigations are needed to investigate whether mir152 might play a role in goblet differentiation or as a tumor suppressor in colon cancer cells.

Results of this study offer a motivation for developing epigenetic therapies that use synthetic miRNAS such as miR ‐152, alone or in combination with other treatments, to reactivate epigenetic silenced genes.

6 CONCLUSIONS AND OUTLOOK

The aim of the present work was the identification of novel genetic factors involved in the differentiation and regeneration of the sensory hair cells in the organ of Corti. In the mammalian auditory system, sensory hair cell loss resulting from aging, ototoxic drugs, infections, overstimulation and other causes is irreversible and leads to permanent sensorineural hearing loss. To restore hearing, it is necessary to generate new functional hair cells. Stem‐cell therapy and molecular therapy are in progress to regenerate hair cells and restore hearing [52]. MicroRNAs emerged as a new class of molecules with potential for gene therapy by taking advantage of their natural role to orchestrate developmental and molecular pathways. MicroRNAs function as master regulators of almost every cellular process where individual miRNAs can repress several hundreds of genes to accomplish biological function.

Besides miRNAs themselves, identification of individual down‐stream targets of miRNAs may reveal novel factors and mechanisms modulating cell fate and regeneration which might lead to new innovative drug or gene therapy strategies.

In order to identify miRNAs with the capability to create sensory hair cells, differential miRNA expression during differentiation of the inner ear progenitor cell line UB/OC1 was analyzed.

Functional characterization of several miRNAs identified by comprehensive small RNA next generation sequencing revealed one candidate, miR‐210, whose knock‐down triggered differentiation from a progenitor cell stage towards a more differentiated hair cell phenotype.

MiR‐210 is described as the “master hypoxamir”, as the induction of miR‐210 is characteristic for the hypoxic response in both normal and transformed cells and is associated with a wide

84 spectrum of miR‐210 targets with roles in mitochondrial metabolism, angiogenesis, DNA repair, and cell survival [81, 111, 112]. Moreover, miR‐210 was found to be increased following erythroid differentiation [113], and it has the ability to induce proliferation of isolated mesenchymal stem cells [114] and to induce angiogenesis and neurogenesis in mouse brain

[115]. It was not identified as being involved in age‐related hearing loss [115] or as being significantly expressed in cochlear sensory epithelia of newborn mice [69]. Since inhibition of miR‐210 in UB/OC‐1 cells changed the cell fate from proliferation to differentiation, we reasoned that miR‐210 plays an active role in maintaining the proliferative progenitor cell stage.

To evaluate the hypothesis that in a reverse situation miR‐210 overexpression may lead to the opposite effect and induce proliferation in differentiated cells, we transduced mouse cochlea with an adenovirus expressing miR‐210. Analysis of the cell fate in a cell lineage tracing model revealed the formation of new hair cells from former Sox2 positive supporting epithelial cells.

The new hair cell formation identified in our model could be possibly due to two mechanisms de‐differentiation or trans‐differentiation. Both mechanisms were previously discussed for sensory hair cell regeneration where transdifferentiation of supporting epithelial cells seems to be the prominent mechanism after forced induction in mammals, for instance by Atoh1, or spontaneously after hair cell damage in non‐mammalian vertebrates [58]. Further in vivo studies would be needed to monitor the regeneration of hair cells in mammalians. One approach would be the delivery of miR‐210 via adenovirus to the deafened auditory epithelium where all hair cells are destroyed, and to analyse the hair cell formation and the auditory brainstem response .

To better understand the mechanisms of new hair cell formation induced by miR‐210 expression, a target prediction analysis using TargetScan was performed and compared to previously published miR‐210 interactions retrieved from Metabase . Our integrative analysis revealed 18 novel candidate targets besides targets predicted previously. To identify candidate genes facilitating miR‐210 mediated transdifferentiation, we performed immunoprecipitation of

RISC complexes enriched for miR‐210 targets followed by quantitative PCR analysis of predicted mouse miR‐210 targets including the targets identified in silico. Besides the identification of several transcripts that are not yet linked to transdifferentiation or sensory epithelium differentiation, we identified several miR‐210 regulated transcripts that are known to play a role in this mechanism.

Of those, brain‐derived neurotrophic factor (Bdnf) is a critical trophic factor required for the development and maintenance of normal innervation of hair cells by afferent spiral ganglion neuron fibers [85] [86]. Changes of the expression levels of Bdnf were also found during transdifferentiation of other cell systems [116], and Bdnf might support transdifferentiation to sensory hair cells once the process is initiated.

Recently, Hoxa1, a member of the homeobox (Hox) transcription factor family regulating embryonic patterning and organogenesis, was found to be transiently expressed in the developing otic epithelium. It is thought to play a role in early regional patterning, thereby contributing to cell lineage development in the inner ear [87]. Microarray analyses of Hoxa1 null embryo mice compared to wild type mice revealed downstream targets of Hoxa1 necessary for early inner ear development, such as fibroblast growth factor receptor‐3 (Fgfr3), which was the only validated downstream target that was up‐regulated in Hoxa1 mutants [88]. Fgfr3 is

86 necessary for the development of the organ of Corti and is known to regulate the differentiation of sensory hair cells and supporting cells. Inhibition of Fgfr in the basilar papilla of birds results in increased hair cell formation and this increase was not associated with increased proliferation, suggesting that inhibition of the Fgf pathway leads to the direct conversion of supporting cells into hair cells [90]. Since loss of FGFR3 leads to excess hair cell development in the mouse organ of Corti [89], transdifferentiation is likely the mechanism in the mammalian auditory sensory epithelium. Thus, inhibition of Fgfr3 via the miR‐210/Hoxa1 pathway might contribute to the transdifferentiation of supporting cells to hair cells that we observed in our experiments.

Another strong candidate for mediating miR‐210 triggered transdifferentiation is Kctd11 (REN).

Kctd11 was previously shown to function as Hedgehog antagonist, playing a role as developmental regulator of neural cell differentiation and regulating proliferation and apoptosis of developing granule cell progenitors. Kctd11 functional knock‐down was shown to impair

Hedgehog antagonism resulting in sustained proliferation of granule progenitor cells, a mechanism responsible for medulloblastoma development [91, 92]. Sonic hedgehog (Shh) signaling is also essential for inner ear sensory epithelia development. In Shh knockout mice, the cochlear sensory organ and spiral ganglion cells are not formed [117], and Shh can promote mouse inner ear progenitor cell proliferation and hair cell differentiation in vitro [118].

Hedgehog signaling was further found to regulate hair cell differentiation in the mammalian cochlea [119]. Moreover, Shh signaling in damaged postnatal rat cochleae renewed proliferation of supporting cells and hair cells, and some proliferating supporting cells are likely to transdifferentiate into hair cells [120]. These findings are in line with our hypothesis that miR‐

210 mediated Kctd11 knock‐down via re‐activated Shh signaling results in transdifferentiation of supporting epithelial cells and new hair cell formation similar as observed by Lu et al. [120].

Deltex‐1 (Dtx1) is described as a transcriptional regulator downstream of the Notch receptor and via inhibition of the transcription factor MASH1 is responsible for the differentiation inhibition of neural progenitor cells [93]. The role of the Notch pathway in inner ear development as well as hair cell regeneration is entirely described [121], making Dtx1 another candidate target for miR‐210 mediated transdifferentiation of supporting epithelial cells to sensory hair cells.

The CRISPR/Cas9 system could be considered to knockout the predicted miR‐210 targets and to investigate their potential in hair cell formation in vitro and in vivo. It was shown recently that

Cas9 nuclease complexed with polyanionic single guide RNA (sgRNA) recognizing a specific sequence of the target can be efficiently delivered into the mammalian inner ear [122].

Taken together, the identification of miR‐210, driving supporting epithelial cells towards the sensory hair cell phenotype, provides new avenues for the treatment of hearing loss. Further validation of down‐stream targets may lead to the discovery of novel drug targets to cure deafness.

In the second part of the thesis, a genetic screen was conducted to identify Atoh1 modulators.

Atoh1 is the mouse homolog of the Drosophila gene atonal that encodes a basic‐helix‐loop‐helix transcription factor. Atoh1 has been shown to act as a “pro‐hair‐cell gene” and is required for the differentiation of hair cells from multipotent progenitors. Experimental overexpression of

Atoh1 in nonsensory cells of the normal organ of Corti generates new hair cells, both in vitro and in vivo [36,37] .

Using a human intestinal epithelial cell line (SW480), engineered with a luciferase reporter gene under the control of the Atoh1 regulatory locus, we screened a library of miRNA expressing plasmids. This led to the identification of miR‐152 acting as a factor leading to reactivation of

Atoh1 gene expression. Reactivation and expression of Atoh1 was associated with reduced DNA methylation of the enhancer region in the Atoh1 locus. These results are consistent with the suggestion that miR‐152 is acting by inhibition of expression of DNA‐methyltransferases

(DNMTs). To further characterize the link between mir152 and Atoh1 activation in the inner ear organ of Corti, miR‐152 was overexpressed in cochlear explants using a viral vector. No activation in the Atoh1 supporting cells was seen in the organ of Corti explants after overexpression of miR‐152. These results provide evidence that miR‐152 activates Atoh1 via epigenetic modulation in the intestinal colon cancer cell line but shows no effect in the reactivation of Atoh1 and in the generation of new hair cells in the organ of Corti. Further investigations would be needed to study whether miR‐152 could play a role in goblet differentiation or as tumor suppressor in colon cancer cells.

In conclusion, the molecular profiling of a precursor inner ear hair cell line provides a tool to identify new genetic factors involved in the differentiation of the sensory hairs cells and may be used to discover new molecular pathways involved in this process. MicroRNAs and the identification of individual down‐stream targets of miRNAs may reveal novel factors and mechanisms modulating cell fate and regeneration, potentially leading to new innovative drug or gene therapy strategies.

7 ACKNOWLEDGMENTS

This PhD work was performed at Novartis Institutes for Biomedical Research (NIBR) in Basel.

I wish to thank Bernd Kinzel my supervisor for giving me the opportunity of doing my PhD in such an interesting field and at such great scientific environment. I’m greatful for the guidance, time and advices for all this time.

I wish to thank Prof. Michael Detmar for the scientific feedback and also for giving me the opportunity of being a PhD student at ETH Zürich. I started at ETH in 2003 as a pharmacist and

I’m honored that I could conclude also my PhD study in such honourable institution.

I also want to thank Prof. Jonathan Hall for accepting to be the co‐referee of this thesis.

For the study: “Mir‐210 promotes ectopic sensory hair cell formation via transdifferentiation of supporting cells in the organ of Corti “ a special thanks goes to the Next Generation Sequencing

Team at Novartis, especially to Guglielmo Roma and Sebastian Bergling for the support in the data analysis, it was great to collaborate with you. A special thanks also to Frederic Sigoillot who helped me with the microRNAs target prediction analysis. I want to thank Prof. Holley from the

University of Sheffield for providing the UB/OC‐1 cell line.

For the study: “Mir‐152 can epigenetically modulate and reactivate the transcription factor

Atoh1”. A special thank goes to Christian Parker and Juliet Davies‐Leighton for the support with the screening and also for the great scientific discussions. Thanks for giving me the possibility to work in your lab, I enjoyed this time a lot.

Immense thanks goes to the wonderful DMP team in the 386 building. All present and former members have not just been my colleagues but also good friends. Especially thanks goes to

Annick Werner, Carole Manneville and Samuel Barbieri, that intensively shared with me Ph.D. happiness and frustration. Thanks to Vanessa, Monika, Patricia, Hyunwoo, Munkyung, Claudia,

Walter, Thomas, Viki, Fede for all the good time spend together.

Finally, I want to thank my family for their endless support during the time of this work.

8 REFERENCES

1. Beattie, J.A., Social aspects of acquired hearing loss in adults. Int J Rehabil Res, 1984. 7(2): p. 215‐ 6. 2. Cohn, E.S. and P.M. Kelley, Clinical phenotype and mutations in connexin 26 (DFNB1/GJB2), the most common cause of childhood hearing loss. Am J Med Genet, 1999. 89(3): p. 130‐6. 3. Andrade, G.M., et al., Hearing loss in congenital toxoplasmosis detected by newborn screening. Braz J Otorhinolaryngol, 2008. 74(1): p. 21‐8. 4. Avettand‐Fenoel, V., et al., Congenital cytomegalovirus is the second most frequent cause of bilateral hearing loss in young French children. J Pediatr, 2013. 162(3): p. 593‐9. 5. Brookhouser, P.E., Sensorineural hearing loss in children. Pediatr Clin North Am, 1996. 43(6): p. 1195‐216. 6. Daniel, E., Noise and hearing loss: a review. J Sch Health, 2007. 77(5): p. 225‐31. 7. Davidson, J., M.L. Hyde, and P.W. Alberti, Epidemiologic patterns in childhood hearing loss: a review. Int J Pediatr Otorhinolaryngol, 1989. 17(3): p. 239‐66. 8. Morata, T.C., Young people: their noise and music exposures and the risk of hearing loss. Int J Audiol, 2007. 46(3): p. 111‐2. 9. Brummett, R.E. and K.E. Fox, Aminoglycoside‐induced hearing loss in humans. Antimicrob Agents Chemother, 1989. 33(6): p. 797‐800. 10. Schacht, J., Molecular mechanisms of drug‐induced hearing loss. Hear Res, 1986. 22: p. 297‐304. 11. Isaacson, J.E. and N.M. Vora, Differential diagnosis and treatment of hearing loss. Am Fam Physician, 2003. 68(6): p. 1125‐32. 12. Staecker, H., Broadening the spectrum of treatment options for SNHL. Arch Otolaryngol Head Neck Surg, 2005. 131(8): p. 734. 13. Torres, M. and F. Giraldez, The development of the vertebrate inner ear. Mech Dev, 1998. 71(1‐ 2): p. 5‐21. 14. Haddon, C., et al., Hair cells without supporting cells: further studies in the ear of the zebrafish mind bomb mutant. J Neurocytol, 1999. 28(10‐11): p. 837‐50. 15. White, P.M., et al., Mammalian cochlear supporting cells can divide and trans‐differentiate into hair cells. Nature, 2006. 441(7096): p. 984‐7. 16. Kelley, M.W., Regulation of cell fate in the sensory epithelia of the inner ear. Nat Rev Neurosci, 2006. 7(11): p. 837‐49. 17. Kiernan, A.E., et al., Sox2 is required for sensory organ development in the mammalian inner ear. Nature, 2005. 434(7036): p. 1031‐5. 18. Lee, Y.S., F. Liu, and N. Segil, A morphogenetic wave of p27Kip1 transcription directs cell cycle exit during organ of Corti development. Development, 2006. 133(15): p. 2817‐26. 19. Wu, D.K. and M.W. Kelley, Molecular mechanisms of inner ear development. Cold Spring Harb Perspect Biol, 2012. 4(8): p. a008409. 20. Mulvaney, J. and A. Dabdoub, Atoh1, an essential transcription factor in neurogenesis and intestinal and inner ear development: function, regulation, and context dependency. J Assoc Res Otolaryngol, 2012. 13(3): p. 281‐93. 21. Bermingham, N.A., et al., Math1: an essential gene for the generation of inner ear hair cells. Science, 1999. 284(5421): p. 1837‐41. 22. Yang, J., et al., Ectopic hair cell‐like cell induction by Math1 mainly involves direct transdifferentiation in neonatal mammalian cochlea. Neurosci Lett, 2013. 549: p. 7‐11.

23. Ahmed, M., et al., Eya1‐Six1 interaction is sufficient to induce hair cell fate in the cochlea by activating Atoh1 expression in cooperation with Sox2. Dev Cell, 2012. 22(2): p. 377‐90. 24. Shi, F., et al., Beta‐catenin up‐regulates Atoh1 expression in neural progenitor cells by interaction with an Atoh1 3' enhancer. J Biol Chem, 2010. 285(1): p. 392‐400. 25. Lanford, P.J., et al., Notch signalling pathway mediates hair cell development in mammalian cochlea. Nat Genet, 1999. 21(3): p. 289‐92. 26. Lanford, P.J., et al., Expression of Math1 and HES5 in the cochleae of wildtype and Jag2 mutant mice. J Assoc Res Otolaryngol, 2000. 1(2): p. 161‐71. 27. Groves, A.K., The challenge of hair cell regeneration. Exp Biol Med (Maywood), 2010. 235(4): p. 434‐46. 28. Mizutari, K., et al., Notch inhibition induces cochlear hair cell regeneration and recovery of hearing after acoustic trauma. Neuron, 2013. 77(1): p. 58‐69. 29. Corwin, J.T., Identifying the genes of hearing, deafness, and dysequilibrium. Proc Natl Acad Sci U S A, 1998. 95(21): p. 12080‐2. 30. Matsui, J.I., et al., Regeneration and replacement in the vertebrate inner ear. Drug Discov Today, 2005. 10(19): p. 1307‐12. 31. Chen, P., et al., The role of Math1 in inner ear development: Uncoupling the establishment of the sensory primordium from hair cell fate determination. Development, 2002. 129(10): p. 2495‐505. 32. Chen, P. and N. Segil, p27(Kip1) links cell proliferation to morphogenesis in the developing organ of Corti. Development, 1999. 126(8): p. 1581‐90. 33. Kanzaki, S., et al., p27(Kip1) deficiency causes organ of Corti pathology and hearing loss. Hear Res, 2006. 214(1‐2): p. 28‐36. 34. Ono, K., et al., Silencing p27 reverses post‐mitotic state of supporting cells in neonatal mouse cochleae. Mol Cell Neurosci, 2009. 42(4): p. 391‐8. 35. Cafaro, J., G.S. Lee, and J.S. Stone, Atoh1 expression defines activated progenitors and differentiating hair cells during avian hair cell regeneration. Dev Dyn, 2007. 236(1): p. 156‐70. 36. Kelley, M.W., D.R. Talreja, and J.T. Corwin, Replacement of hair cells after laser microbeam irradiation in cultured organs of corti from embryonic and neonatal mice. J Neurosci, 1995. 15(4): p. 3013‐26. 37. Shou, J., J.L. Zheng, and W.Q. Gao, Robust generation of new hair cells in the mature mammalian inner ear by adenoviral expression of Hath1. Mol Cell Neurosci, 2003. 23(2): p. 169‐79. 38. Atkinson, P.J., et al., Hair cell regeneration after ATOH1 gene therapy in the cochlea of profoundly deaf adult guinea pigs. PLoS One, 2014. 9(7): p. e102077. 39. Ha, M. and V.N. Kim, Regulation of microRNA biogenesis. Nat Rev Mol Cell Biol, 2014. 15(8): p. 509‐24. 40. Zhang, B., Q. Wang, and X. Pan, MicroRNAs and their regulatory roles in animals and plants. J Cell Physiol, 2007. 210(2): p. 279‐89. 41. Sivakumaran, T.A., et al., Characterization of an abundant COL9A1 transcript in the cochlea with a novel 3' UTR: Expression studies and detection of miRNA target sequence. J Assoc Res Otolaryngol, 2006. 7(2): p. 160‐72. 42. Weston, M.D., et al., MicroRNA gene expression in the mouse inner ear. Brain Res, 2006. 1111(1): p. 95‐104. 43. Wienholds, E. and R.H. Plasterk, MicroRNA function in animal development. FEBS Lett, 2005. 579(26): p. 5911‐22. 44. Mencia, A., et al., Mutations in the seed region of human miR‐96 are responsible for nonsyndromic progressive hearing loss. Nat Genet, 2009. 41(5): p. 609‐13. 45. Friedland, D.R., et al., Cholesteatoma growth and proliferation: posttranscriptional regulation by microRNA‐21. Otol Neurotol, 2009. 30(7): p. 998‐1005.

46. Friedman, L.M., et al., MicroRNAs are essential for development and function of inner ear hair cells in vertebrates. Proc Natl Acad Sci U S A, 2009. 106(19): p. 7915‐20. 47. Li, H. and D.M. Fekete, MicroRNAs in hair cell development and deafness. Curr Opin Otolaryngol Head Neck Surg, 2010. 18(5): p. 459‐65. 48. Barald, K.F., et al., Immortalized cell lines from embryonic avian and murine otocysts: tools for molecular studies of the developing inner ear. Int J Dev Neurosci, 1997. 15(4‐5): p. 523‐40. 49. Rivolta, M.N., et al., Auditory hair cell precursors immortalized from the mammalian inner ear. Proc Biol Sci, 1998. 265(1406): p. 1595‐603. 50. Jat, P.S., et al., Direct derivation of conditionally immortal cell lines from an H‐2Kb‐tsA58 transgenic mouse. Proc Natl Acad Sci U S A, 1991. 88(12): p. 5096‐100. 51. Rivolta, M.N. and M.C. Holley, Cell lines in inner ear research. J Neurobiol, 2002. 53(2): p. 306‐18. 52. Geleoc, G.S. and J.R. Holt, Sound strategies for hearing restoration. Science, 2014. 344(6184): p. 1241062. 53. Zilberstein, Y., M.C. Liberman, and G. Corfas, Inner hair cells are not required for survival of spiral ganglion neurons in the adult cochlea. J Neurosci, 2012. 32(2): p. 405‐10. 54. Cruz, R.M., P.R. Lambert, and E.W. Rubel, Light microscopic evidence of hair cell regeneration after gentamicin toxicity in chick cochlea. Arch Otolaryngol Head Neck Surg, 1987. 113(10): p. 1058‐62. 55. Corwin, J.T. and D.A. Cotanche, Regeneration of sensory hair cells after acoustic trauma. Science, 1988. 240(4860): p. 1772‐4. 56. Jones, J.E. and J.T. Corwin, Regeneration of sensory cells after laser ablation in the lateral line system: hair cell lineage and macrophage behavior revealed by time‐lapse video microscopy. J Neurosci, 1996. 16(2): p. 649‐62. 57. Izumikawa, M., et al., Auditory hair cell replacement and hearing improvement by Atoh1 gene therapy in deaf mammals. Nat Med, 2005. 11(3): p. 271‐6. 58. Brigande, J.V. and S. Heller, Quo vadis, hair cell regeneration? Nat Neurosci, 2009. 12(6): p. 679‐ 85. 59. Edge, A.S. and Z.Y. Chen, Hair cell regeneration. Curr Opin Neurobiol, 2008. 18(4): p. 377‐82. 60. Lowenheim, H., et al., Gene disruption of p27(Kip1) allows cell proliferation in the postnatal and adult organ of corti. Proc Natl Acad Sci U S A, 1999. 96(7): p. 4084‐8. 61. Oesterle, E.C., et al., p27(Kip1) is required to maintain proliferative quiescence in the adult cochlea and pituitary. Cell Cycle, 2011. 10(8): p. 1237‐48. 62. Maass, J.C., et al., p27Kip1 knockdown induces proliferation in the organ of Corti in culture after efficient shRNA lentiviral transduction. J Assoc Res Otolaryngol, 2013. 14(4): p. 495‐508. 63. Petersen, C.P., et al., Short RNAs repress translation after initiation in mammalian cells. Mol Cell, 2006. 21(4): p. 533‐42. 64. Guo, H., et al., Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature, 2010. 466(7308): p. 835‐40. 65. Li, Z. and T.M. Rana, Therapeutic targeting of microRNAs: current status and future challenges. Nat Rev Drug Discov, 2014. 13(8): p. 622‐38. 66. Wienholds, E., et al., MicroRNA expression in zebrafish embryonic development. Science, 2005. 309(5732): p. 310‐1. 67. Rudnicki, A. and K.B. Avraham, microRNAs: the art of silencing in the ear. EMBO Mol Med, 2012. 4(9): p. 849‐59. 68. Rudnicki, A., et al., Next‐generation sequencing of small RNAs from inner ear sensory epithelium identifies microRNAs and defines regulatory pathways. BMC Genomics, 2014. 15: p. 484. 69. Zhang, Y., et al., Isolation, growth and differentiation of hair cell progenitors from the newborn rat cochlear greater epithelial ridge. J Neurosci Methods, 2007. 164(2): p. 271‐9.

70. Kozomara, A. and S. Griffiths‐Jones, miRBase: annotating high confidence microRNAs using deep sequencing data. Nucleic Acids Res, 2014. 42(Database issue): p. D68‐73. 71. Langmead, B., et al., Ultrafast and memory‐efficient alignment of short DNA sequences to the human genome. Genome Biol, 2009. 10(3): p. R25. 72. Anders, S. and W. Huber, Differential expression analysis for sequence count data. Genome Biol, 2010. 11(10): p. R106. 73. Parker, M., A. Brugeaud, and A.S. Edge, Primary culture and plasmid electroporation of the murine organ of Corti. J Vis Exp, 2010(36). 74. Griffiths‐Jones, S., miRBase: the microRNA sequence database. Methods Mol Biol, 2006. 342: p. 129‐38. 75. Griffiths‐Jones, S., et al., miRBase: tools for microRNA genomics. Nucleic Acids Res, 2008. 36(Database issue): p. D154‐8. 76. Xiang, M., et al., Requirement for Brn‐3c in maturation and survival, but not in fate determination of inner ear hair cells. Development, 1998. 125(20): p. 3935‐46. 77. Kanzaki, S., et al., Transgene expression in neonatal mouse inner ear explants mediated by first and advanced generation adenovirus vectors. Hear Res, 2002. 169(1‐2): p. 112‐20. 78. He, J., et al., MiR‐210 disturbs mitotic progression through regulating a group of mitosis‐related genes. Nucleic Acids Res, 2013. 41(1): p. 498‐508. 79. Wang, H., et al., Negative regulation of Hif1a expression and TH17 differentiation by the hypoxia‐ regulated microRNA miR‐210. Nat Immunol, 2014. 15(4): p. 393‐401. 80. Xiang, M., Requirement for Brn‐3b in early differentiation of postmitotic retinal ganglion cell precursors. Dev Biol, 1998. 197(2): p. 155‐69. 81. Devlin, C., et al., miR‐210: More than a silent player in hypoxia. IUBMB Life, 2011. 63(2): p. 94‐ 100. 82. Huang, X., Q.T. Le, and A.J. Giaccia, MiR‐210‐‐micromanager of the hypoxia pathway. Trends Mol Med, 2010. 16(5): p. 230‐7. 83. Zhang, Z., et al., MicroRNA miR‐210 modulates cellular response to hypoxia through the MYC antagonist MNT. Cell Cycle, 2009. 8(17): p. 2756‐68. 84. Qu, A., et al., Hypoxia‐inducible MiR‐210 is an independent prognostic factor and contributes to metastasis in colorectal cancer. PLoS One, 2014. 9(3): p. e90952. 85. Fukui, H., et al., BDNF gene therapy induces auditory nerve survival and fiber sprouting in deaf Pou4f3 mutant mice. Sci Rep, 2012. 2: p. 838. 86. Yang, T., et al., The molecular basis of making spiral ganglion neurons and connecting them to hair cells of the organ of Corti. Hear Res, 2011. 278(1‐2): p. 21‐33. 87. Makki, N. and M.R. Capecchi, Hoxa1 lineage tracing indicates a direct role for Hoxa1 in the development of the inner ear, the heart, and the third rhombomere. Dev Biol, 2010. 341(2): p. 499‐509. 88. Makki, N. and M.R. Capecchi, Identification of novel Hoxa1 downstream targets regulating hindbrain, neural crest and inner ear development. Dev Biol, 2011. 357(2): p. 295‐304. 89. Hayashi, T., D. Cunningham, and O. Bermingham‐McDonogh, Loss of Fgfr3 leads to excess hair cell development in the mouse organ of Corti. Dev Dyn, 2007. 236(2): p. 525‐33. 90. Jacques, B.E., A. Dabdoub, and M.W. Kelley, Fgf signaling regulates development and transdifferentiation of hair cells and supporting cells in the basilar papilla. Hear Res, 2012. 289(1‐ 2): p. 27‐39. 91. Gallo, R., et al., REN: a novel, developmentally regulated gene that promotes neural cell differentiation. J Cell Biol, 2002. 158(4): p. 731‐40. 92. Argenti, B., et al., Hedgehog antagonist REN(KCTD11) regulates proliferation and apoptosis of developing granule cell progenitors. J Neurosci, 2005. 25(36): p. 8338‐46.

93. Yamamoto, N., et al., Role of Deltex‐1 as a transcriptional regulator downstream of the Notch receptor. J Biol Chem, 2001. 276(48): p. 45031‐40. 94. Agrawal, Y., E.A. Platz, and J.K. Niparko, Prevalence of hearing loss and differences by demographic characteristics among US adults: data from the National Health and Nutrition Examination Survey, 1999‐2004. Arch Intern Med, 2008. 168(14): p. 1522‐30. 95. Deltenre, P. and L. Van Maldergem, Hearing loss and deafness in the pediatric population: causes, diagnosis, and rehabilitation. Handb Clin Neurol, 2013. 113: p. 1527‐38. 96. Cho, S.I., et al., Mechanisms of hearing loss after blast injury to the ear. PLoS One, 2013. 8(7): p. e67618. 97. Dror, A.A. and K.B. Avraham, Hearing impairment: a panoply of genes and functions. Neuron, 2010. 68(2): p. 293‐308. 98. Akazawa, C., et al., A mammalian helix‐loop‐helix factor structurally related to the product of Drosophila proneural gene atonal is a positive transcriptional regulator expressed in the developing nervous system. J Biol Chem, 1995. 270(15): p. 8730‐8. 99. Cotanche, D.A. and C.L. Kaiser, Hair cell fate decisions in cochlear development and regeneration. Hear Res, 2010. 266(1‐2): p. 18‐25. 100. Du, X., et al., Wild‐type cells rescue genotypically Math1‐null hair cells in the inner ears of chimeric mice. Dev Biol, 2007. 305(2): p. 430‐8. 101. Klisch, T.J., et al., In vivo Atoh1 targetome reveals how a proneural transcription factor regulates cerebellar development. Proc Natl Acad Sci U S A, 2011. 108(8): p. 3288‐93. 102. Yang, H., et al., Generation and characterization of Atoh1‐Cre knock‐in mouse line. Genesis, 2010. 48(6): p. 407‐13. 103. VanDussen, K.L. and L.C. Samuelson, Mouse atonal homolog 1 directs intestinal progenitors to secretory cell rather than absorptive cell fate. Dev Biol, 2010. 346(2): p. 215‐23. 104. Garzon, R., G. Marcucci, and C.M. Croce, Targeting microRNAs in cancer: rationale, strategies and challenges. Nat Rev Drug Discov, 2010. 9(10): p. 775‐89. 105. Ling, H., M. Fabbri, and G.A. Calin, MicroRNAs and other non‐coding RNAs as targets for anticancer drug development. Nat Rev Drug Discov, 2013. 12(11): p. 847‐65. 106. Xu, Q., et al., A regulatory circuit of miR‐148a/152 and DNMT1 in modulating cell transformation and tumor angiogenesis through IGF‐IR and IRS1. J Mol Cell Biol, 2013. 5(1): p. 3‐13. 107. Ji, W., et al., MicroRNA‐152 targets DNA methyltransferase 1 in NiS‐transformed cells via a feedback mechanism. Carcinogenesis, 2013. 34(2): p. 446‐53. 108. Tsuruta, T., et al., miR‐152 is a tumor suppressor microRNA that is silenced by DNA hypermethylation in endometrial cancer. Cancer Res, 2011. 71(20): p. 6450‐62. 109. Kazanjian, A. and N.F. Shroyer, NOTCH Signaling and ATOH1 in Colorectal Cancers. Curr Colorectal Cancer Rep, 2011. 7(2): p. 121‐127. 110. Bossuyt, W., et al., Atonal homolog 1 is a tumor suppressor gene. PLoS Biol, 2009. 7(2): p. e39. 111. Corn, P.G., Hypoxic regulation of miR‐210: shrinking targets expand HIF‐1's influence. Cancer Biol Ther, 2008. 7(2): p. 265‐7. 112. Chan, Y.C., et al., miR‐210: the master hypoxamir. Microcirculation, 2012. 19(3): p. 215‐23. 113. Bianchi, N., et al., Involvement of miRNA in erythroid differentiation. Epigenomics, 2012. 4(1): p. 51‐65. 114. Minayi, N., et al., The Effect of miR‐210 Up‐regulation on Proliferation and Survival of Mouse Bone Marrow Derived Mesenchymal Stem Cell. Int J Hematol Oncol Stem Cell Res, 2014. 8(1): p. 15‐23. 115. Zeng, L., et al., MicroRNA‐210 overexpression induces angiogenesis and neurogenesis in the normal adult mouse brain. Gene Ther, 2014. 21(1): p. 37‐43.

116. Esmaeili, A., et al., Messenger RNA Expression Patterns of Neurotrophins during Transdifferentiation of Stem Cells from Human‐Exfoliated Deciduous Teeth into Neural‐like Cells. Avicenna J Med Biotechnol, 2014. 6(1): p. 21‐6. 117. Riccomagno, M.M., et al., Specification of the mammalian cochlea is dependent on Sonic hedgehog. Genes Dev, 2002. 16(18): p. 2365‐78. 118. Zhao, Y., et al., Sonic hedgehog promotes mouse inner ear progenitor cell proliferation and hair cell generation in vitro. Neuroreport, 2006. 17(2): p. 121‐4. 119. Tateya, T., et al., Hedgehog signaling regulates prosensory cell properties during the basal‐to‐ apical wave of hair cell differentiation in the mammalian cochlea. Development, 2013. 140(18): p. 3848‐57. 120. Lu, N., et al., Sonic hedgehog initiates cochlear hair cell regeneration through downregulation of retinoblastoma protein. Biochem Biophys Res Commun, 2013. 430(2): p. 700‐5. 121. Murata, J., K. Ikeda, and H. Okano, Notch signaling and the developing inner ear. Adv Exp Med Biol, 2012. 727: p. 161‐73. 122. Zuris, J.A., et al., Cationic lipid‐mediated delivery of proteins enables efficient protein‐based genome editing in vitro and in vivo. Nat Biotechnol, 2015. 33(1): p. 73‐80.