Using Sox2 to alleviate the hallmarks of age-related hearing loss

Abstract

Age-related hearing loss (ARHL) is the most prevalent age-related sensory deficit. ARHL reduces the quality of life of the growing aging population, setting seniors up for the enhanced mental decline. The size of the needy population, the structural deficit, and a likely research strategy for effective treatment of chronic neurosensory hearing in the elderly are needed. Although there has been profound advancement in auditory regenerative research, there remain multiple challenges to restore hearing loss. Thus, additional investigations are required, using novel tools. We propose how the flat epithelium, remaining after the organ of Corti has deteriorated, can be converted to the repaired-sensory epithelium, using Sox2. This will include developing an artificial gene regulatory network transmitted by large viral vectors to the flat epithelium to stimulate remnants of the organ of Corti to restore hair cells. We hope to unite with our proposal toward the common goal, eventually restoring a functional human hearing organ by transforming the flat epithelial cells left after the organ of Corti loss.

Graphical abstract

1. Introduction

Progressive age-related hearing loss (ARHL) of hair cells (https://www.who.int/news-room/fact-sheets/detail/deafness-and-hearing-loss) is the most profound sensory loss that will reach about ~2 billion older adults (Hoffman et al., 2017; Schilder et al., 2019). Without a doubt, the detailed cellular architecture of the organ of Corti is essential for the function of the mammalian hearing organ to transduce sound into electric signals effectively (Ashmore, 2018; Jahan et al., 2015c; Tan et al., 2018; Xia et al., 2018). While the basic helix-loop-helix (bHLH) transcription factor Atoh1 alone, or in combination with other transcription factors, can induce transformation of supporting cells into hair cells (Atkinson et al., 2019), the process mostly stalls after day 14 in postnatal mice (Driver et al., 2017; Kelly et al., 2012) and multiple co-factors are needed to induce limited hair cell transformation in older mice (McGovern et al., 2019; Yamashita et al., 2018). In addition to the diminished success rate of such conversion with age, converting supporting cells to hair cells will remove functionally important supporting cells (Zine et al., 2014). Unfortunately, embryonic proliferation in the organ of Corti (Matei et al., 2005; Ruben, 1967) comes to a halt shortly after the onset of hearing (White et al., 2006). Even radical changes in the cell-cycle regulation have limited effects in terms of jump-starting the proliferation (Kopecky et al., 2013; Mantela et al., 2005; Schimmang and Pirvola, 2013; Silva and Maass, 2019; Zhang et al., 2018), possibly because of depletion of endogenous stem cells (Lenz et al., 2019; Lopez-Juarez et al., 2019).

Interactions of hair cells with the surrounding supporting cells are required to form the functionally essential cellular assembly in the inner ear (Fritzsch et al., 2006; Groves and Fekete, 2012, 2017; Petrovic et al., 2015). Among the transcription factors are Eya1/Six1 (Ahmed et al., 2012; Li et al., 2020; Zou et al., 2004), Sox2 (Dvorakova et al., 2016; Dvorakova et al., 2020; Kiernan et al., 2005), Gata3 (Duncan and Fritzsch, 2013; Karis et al., 2001; Walters et al., 2017), Pax2 (Bouchard et al., 2010; Kempfle and Edge, 2014) and several Lim domain factors such as Lmx1a (Huang et al., 2008; Huang et al., 2018; Nichols et al., 2020; Nichols et al., 2008), Isl1 (Huang et al., 2008) and Lhx3 (Hertzano et al., 2007). How all these factors interact to allow, for example, a very tight Atoh1 expression regulation (Cheng et al., 2016; Costa et al., 2017; Jahan et al., 2015b) in cyclical sequential fashion necessary for healthy sensory neuronal development (Fritzsch et al., 2006; Kageyama et al., 2018; Tateya et al., 2019) remains unclear. Once expressed, Atoh1 induces the transformation of naive postmitotic epithelial cells or supporting cells into hair cells, but the viability of those cells is limited (Atkinson et al., 2019; Jahan et al., 2018) beyond neonatal transformation (Kelly et al., 2012; Koehler et al., 2017; McGovern et al., 2019). Currently, most mouse models for transformation focus on acute hair cell loss (Li et al., 2020; Roccio et al., 2019; Walters et al., 2017; Yamashita et al., 2018), whereas age-related hearing loss is much less explored (Devare et al., 2018; Zine et al., 2014). Importantly, the stages of mouse development are not equivalent to humans (Lim and Brichta, 2016; Wang et al., 2018), and two-year-old mice are the equivalent of ~70-year-old humans (Fig. 1). The human-age at which hair cell and spiral ganglion neuron loss become particularly profound (Eggermont, 2019; Kusunoki et al., 2004; Rauch et al., 2001; Sheffield and Smith, 2018) are dissimilar to mice. In fact, hardly any restorative work has been conducted in mice over two years of age. It is clear that current literature suggests that human hair cells and spiral ganglion neurons loss with age are not necessarily in synchrony with each other (Eggermont, 2019; Liberman, 2017; Makary et al., 2011; Schuknecht and Gacek, 1993).

Fig. 1.

Establishing a mouse model to test hearing restoration requires an understanding of the equivalent ages of mice and humans (blue curve in a). Ear development starts with the invaginating otic placode in mice and humans. Human hearing onset is around gestational week 19 (b), whereas the beginning of hearing in mice is around 12 days after birth (a). Note that the mouse develops over 100 times faster up to 1 month after birth but progressively slows down with age to about 25 times faster (b). A 70-year-old human is approximately equivalent to a 2-year-old mouse. It is the age group in humans that shows the most profound hearing decline, displayed as a percent (%) of hearing loss (red curve in a) resulting in >25 dB hearing loss in ~90% of 90-year-olds. Importantly, hardly any restoration attempts have been conducted on 2-year-old mice due to limited viability, cost of keeping animals to old age, and lack of models. Data were taken from the Jackson lab, WHF, and (Sheffield and Smith, 2018).

The focus on early postnatal and acute hair cell loss is in part related to the progressive deterioration of the organ of Corti, including loss of supporting cells after long-term hair cell loss (Herranen et al., 2020; Kersigo and Fritzsch, 2015; Pauley et al., 2008; Taylor et al., 2012), resulting in a flat epithelium (Shibata et al., 2010). Such ‘flat epithelium,’ combined with variable sensory neuron loss, maybe the prevailing condition encountered in human age-related hearing loss (Atkinson et al., 2019; Liberman, 2017; Lopez-Juarez et al., 2019), which constitute the vast majority of hearing deficient people requiring a cure (Dawson and Bowl, 2018; Dubno, 2019; Homans et al., 2017; Lopez-Juarez et al., 2019; Revuelta et al., 2017; Yamasoba et al., 2013; Yang et al., 2015). Indeed, numerous studies and companies focus on various aspects of age-related hearing prevention and multiple approaches toward curing age-related hearing loss (Eshraghi et al., 2019; Schilder et al., 2019). Integrated with these approaches are more recent attempts to define the genetic basis of ARHL (Lewis et al., 2018a) and improve measuring hearing loss progression (Cassarly et al., 2019; Simpson et al., 2019) to guide attempts to restore hair cells and hearing.

The purpose of the current review is to explore existing, partially described mouse mutants to promote a genetic model with at least a somewhat flat epithelium as a phenotype to cure now defined hallmarks of ARHL, the ‘flat epithelium.’ We will describe such a mouse model, Sox2^Ysb, to provide better integration of current attempts of regeneration into genetically-defined mouse lines. We hope that this approach will fuel, in analogy to the stimulation achieved by formulating molecular hallmarks of aging (Campisi, 2013; López-Otín et al., 2013), improved future translatability of current restoration attempts. Keeping the proposed novel mouse model’s strengths and weaknesses in mind, we suggest a series of experiments that could allow flat epithelial cells to respond to molecular treatment with hair cell differentiation and perhaps reconstitution of the organ of Corti. It would be extremely beneficial to develop restoration strategies in a mouse model that is temporally scalable from postnatal to old age to minimize breeding costs. While age-dependent dysregulation of endolymph composition (Dubno, 2019) and neuronal loss (Lang, 2016) and their potential restoration for hearing improvement are being explored (Fritzsch et al., 2019), we focus here on hair cell and organ of Corti regeneration models by first defining the magnitude of ARHL, secondly, determining the direction to follow and use for regeneration, and thirdly proposing a novel approach to combine Sox2 and Atoh1 as potential strategy to reestablish hearing. We will first provide a background on current attempts to restore hearing loss and the problems faced. We follow this by recommending a new approach and future direction to use large viral vectors to transfect flat epithelial cells in a hearing loss mouse model with a construct to drive gene regulatory networks needed for hair cell development.

2. Defining the magnitude of the problem of age-related hearing loss

Congenital hearing impairment has a prevalence of ~3 per 1000 live birth (~ 2 million of the ~700 million babies born in 2017, according to http://worldpopulationreview.com). Human hearing loss has three major causes:

1. Hearing loss due to genetic defects (mostly explored as congenital hearing loss)

2. Hearing loss induced by loud sound or ototoxic drug exposure (induced hearing loss)

3. Age-related hearing loss (due to genetics, environment, and aging).

Of these, ~ 60% are deaf due to inherited causes (Devare et al., 2018; Eshraghi et al., 2019; Korver et al., 2017; Mizutari et al., 2015) with ionic balance disorders being the most common forms. Unfortunately, limited progress has been made to combat the most widely distributed congenital hearing loss, mutations in Connexin 26 (GJB2) that is involved in K⁺ recirculation (estimated to be nearly 50% in some western populations) or Pendrin (Slc26a4) mutation that appears to be around 40% in some Asian communities (Sheffield and Smith, 2018; Yu et al., 2018).

In contrast, significant progress has been made to combat some forms of congenital hearing loss caused by single-gene mutations (Erives and Fritzsch, 2020; Géléoc and Holt, 2014; Lustig and Akil, 2018; Shibata et al., 2016). Progress has also been made to combat noise-induced hearing loss by using appropriate protective gear. Furthermore, by regulating levels of ototoxic drug treatment, such as certain aminoglycosides or cisplatin, the damaging effects on the hearing can be counteracted (Roccio et al., 2019; Schilder et al., 2019; Yang et al., 2015). As we better understand how to avoid negative consequences of antibiotic and anti-cancer treatments, some of their effects on acquired hearing loss may be further reduced (Dawson and Bowl, 2018; Schilder et al., 2019; Sheffield and Smith, 2018).

Despite this progress concerning congenital and acquired hearing loss, the growing problem of ARHL is only now coming into full focus (Dawson and Bowl, 2018; Eshraghi et al., 2019; Revuelta et al., 2017; Schilder et al., 2019). Recent World health organization (WHO) estimates assume that by 2050, the number of people 65 years or older will reach about ~2 billion, primarily fueled by the baby boomer cohort in combination with the increased life span of 80 years or more in the developed countries. While the prevalence of hearing loss is likely to be substantial, the real magnitude of the problem is difficult to estimate as it constitutes a combination of conductive hearing loss (middle ear defects) with neurosensory hearing loss (for various reasons including loss of hair cells or neurons) (Dubno, 2019; Fattal et al., 2018; Kusunoki et al., 2004; Lang, 2016; Makary et al., 2011; Schuknecht and Gacek, 1993; Sheffield and Smith, 2018). Since about 31% of women and about 33% of men over age 65 have a hearing sensitivity loss of 35 dB or more (Eggermont, 2019; Homans et al., 2017), would indicate that roughly 30% of the approximately 2 billion people over 65 years of age (about 600 million people worldwide) may have progressive neurosensory hearing loss. Current estimates also imply that roughly 90% of people aged 90 and older will suffer from some form of hearing decline in at least one ear (Fig. 1). Using mice models with hearing phenotype akin to humans would involve undue cost on keeping mice. Alternatively, even if mouse strains with premature hearing loss are used (Ohlemiller et al., 2016; Zheng et al., 1999), it pushes the viability near the 50% mortality line established for mice (https://www.jax.org/research-and-faculty/research-labs/the-harrison-lab/gerontology/life-span-as-a-biomarker).

More recently, over approximately 300 gene variants have been identified that are associated with ARHL (Bowl et al., 2017). However, substantial cohort evidence has dramatically reduced the likely effect of genetic predisposition on longevity in general to under 20% (Kaplanis et al., 2018). This raises the possibility that a better understanding of the genetic predisposition of ARHL (Lewis et al., 2018a) will benefit only about 1/5 of those affected. Part of the problem in defining ARHL and the size of its genetic basis relates to the difficulty in estimating the compounding impact of loud sound or of chemical exposure on humans (Bielefeld et al., 2010; Liberman, 2017; Yang et al., 2015), a problem that may even be more impacted by the use of earphones in the younger generations. Given all of these uncontrollable variables, we want to focus here on problems related to defining controllable mouse models of ARHL. Specifically, we propose a novel mouse model, Sox2^Ysb, that allows evaluating restoration successfully in terms of functional assessment of such restoration attempts at any time, ranging from early postnatal to senescence.

3. Problems compounding directions to generate a novel hearing loss

As is evident from the outline, the most common hearing problem, ARHL, will benefit little from currently championed approaches in mice (Fig. 1). This is so because most strategies mimic the direct transformation of supporting cells into hair cells using the single-gene expression approach introduced by Weintraub 30 years ago (Weintraub et al., 1989). However, questions after this breakthrough (Murre, 2019) and even in systems that are much simpler than the mammalian organ of Corti (Jahan et al., 2018; Jahan et al., 2015a) remain unclear. To achieve the reported effects of direct supporting cell transformation, most attempts focus on transforming cells in an already well established early postnatal scaffold (Atkinson et al., 2019). However, regeneration is complicated by the fact that the organ of Corti requires a sophisticated, stereotyped cellular assembly (Tan et al., 2018) to function appropriately in transforming sound energy effectively into electric signals. Such morphological details involve highly ordered organization of the cellular assembly (Jahan et al., 2015a), and precise alignment of the tectorial membrane with the stereocilia of the hair cells (Elliott et al., 2018; Goodyear and Richardson, 2018; Nist-Lund et al., 2019) to translate the limited movements of the organ of Corti during normal sound levels (Fettiplace, 2017; Ren et al., 2016).

The alternative to direct transformation of cells using modifications of the Weintraub approach is to use the Yamanaka approach of reprogramming (Takahashi and Yamanaka, 2006) to induce pluripotency in the dish (Oct3/4, Sox2, c-Myc, and Kif4), followed by transdifferentiation (Riemens et al., 2018) and insertion of immature hair cells into the scala media (Lopez-Juarez et al., 2019) without inducing tumor pathologies (Cutfield et al., 2019). Progress has been made to generate vestibular and cochlear hair cells out of induced pluripotent human stem cells (Koehler et al., 2017; Lahlou et al., 2018; Patel et al., 2018). However, thus far, no one has succeeded in generating both types of the organ of Corti hair cells in vitro in the right proportion, and in differential distribution needed for healthy organ of Corti restoration (Elliott et al., 2018). For many of these issues, we have no solution at hand, turning the feasibility of generating hair cells with subsequent repopulation of the organ of Corti into a treatment option of the future, after those issues are resolved.

The development of the healthy organ of Corti is susceptible to variations of gene signaling at different levels (Jahan et al., 2015c), including regulatory interactions of micro RNAs and in RNAs (Booth et al., 2018). These add yet another gene regulatory mechanism that complicates hearing restoration. Moreover, very few mutants have thus far been characterized that show a patchy loss of organ of Corti and are viable (Kiernan et al., 2005; Pirvola et al., 2002; Soukup et al., 2009), more comparable to what clinicians encounter in the aging cochlea (Kusunoki et al., 2004; Liberman, 2017; Yamasoba et al., 2013). Such models would allow monitoring not only the restoration of the flat epithelium between the remaining patches of hair cells (Duncan and Fritzsch, 2013; Herranen et al., 2020; Kiernan et al., 2005; Pan et al., 2012; Pirvola et al., 2002; Soukup et al., 2009) but also the restoration of innervation to the newly formed hair cells.

4. Toward defining an ideal mouse model for organ of Corti restoration

We focus here on mice that have mutations in enhancer regions causing partial or complete loss of hair cell differentiation due to lack of or limited expression of Sox2, the light coat and circling (Lcc), and yellow submarine (Ysb) mutants (Kiernan et al., 2005). In particular, the Sox2^Ysb mutation shows patchy hair cell development in some areas of the organ of Corti, interconnected by flat epithelium (Fig. 2). We recently investigated the pattern of innervation in the Ysb mutant and found, consistent with data in other mutants with a partial loss of hair cells (Jahan et al., 2018; Jahan et al., 2015c; Tan et al., 2018), that radial fibers turn toward the remaining patches of hair cells (Fig. 2) possibly attracted by trophic factors released from the hair cells (Tessarollo et al., 2004). This implies that many spiral ganglion neurons can survive the absence of a direct Organ of Corti target in their trajectory by rerouting to remaining targets as also demonstrated in other Sox2 deletion models (Dvorakova et al., 2016; Dvorakova et al., 2020; Steevens et al., 2019). Restoring the organ of Corti to replace the flat epithelium between remaining patches of hair cells could thus be revealed by the altered trajectories of nerve fibers to demonstrate that re-innervation of replaced hair cells has occurred (Fig. 2). This underexplored mutant could thus become an ideal model to establish the ability to restore the flat epithelium into a functional and continuous organ of Corti, complete with the restoration of a near-normal pattern of innervation at various stages in postnatal development. For obvious reasons, attempts should first start with neonatal animals to verify how much of the established molecular transformation of supporting cells (Walters et al., 2017; Yamashita et al., 2018) or information gained through hair cell formation in vitro (Koehler et al., 2017; McLean et al., 2017; Patel et al., 2018) can be used to fill in the gaps between the patches to regenerate a continuous organ of Corti epithelium in neonatal Ysb mutants. The approach could be refined to earlier stages, including senescent mice of ~2 years of age, to mimic human seniors with long-lasting hearing loss.

Fig. 2.

A comparison of OC organization in E18.5 control (A, A’) and Ysb mutant (B,B’). Note that the continuous distribution of hair cells and supporting cells are changed into patches of sensory epithelia with partially distinct hair cells and innervation by radial fibers. The idea proposed here is to use a sensory patch generating mutant, such as Ysb, to fine-tune viral transfection approaches to turn the BMP4 positive flat epithelium into hair cells that fill in the gap between patches (C, C’). As new hair cells form (green ovals in C’), they will express neurotrophins that can attract fibers to grow toward them (blue lines in C’). The deviation in the fiber trajectory should allow to illustrate the progress of regeneration while fluorescent markers will identify the newly formed hair cells after transfection of flat epithelial cells with a viral vector that holds all regulatory elements to initiate and eventually stop expression of genes necessary for hair cell specification and differentiation (see Fig. 3).

4.1. Proposed approaches and strategies: current understanding

Previous work has established that only two approaches to restoring hearing are potentially viable: Direct cellular transformation with or without the proliferation of precursors and in vitro generation of hair cells followed by transplantation (Atkinson et al., 2019; Roccio et al., 2019; Zine et al., 2014). The later approach has numerous unsolved issues, such as the generation of vestibular instead of cochlear hair cells (Elliott et al., 2018) or only one type of cochlear-like hair cells (Lenz et al., 2019; McLean et al., 2017). Also, no clear strategy exists on how to overcome the toxicity of the endolymph (Park et al., 2014) for in vitro approaches, despite their merit in terms of defining the molecular guidance of hair cell differentiation (Koehler et al., 2017; Patel et al., 2018). It should be considered here that despite recent limited success using human stem cell injections into the perilymphatic spaces (Lopez-Juarez et al., 2019), there are multiple unknowns regarding this strategy. The proliferation of undifferentiated cells or supporting cells followed by transdifferentiation is only now coming into focus (Lopez-Juarez et al., 2019; Roccio and Edge, 2019; Zine et al., 2014) and may hold the potential to replace large areas of the lost organ of Corti above and beyond the limited transformation of a few supporting cells shortly after hair cells loss (McGovern et al., 2019).

Focusing on the proliferation/transdifferentiation of undifferentiated epithelial cells of the flat epithelium, and arguing that multiple steps need to be taken before such precursors can respond to Sox2 and Atoh1 (Dabdoub et al., 2008; Dvorakova et al., 2020) with appropriate hair cell differentiation in the aging cochlea are prime objectives. A promising idea is an approach successfully employed for intestinal cell repopulation using Lgr5 positive cells to generate both in vitro and in vivo intestinal epithelium (Gehart and Clevers, 2018; Li et al., 2019). While approaches inspired by the gut epithelium Lgr5-positive cell expansion seem to be somewhat successful in generating hair cells in vitro (Lenz et al., 2019; McLean et al., 2017; Roccio et al., 2018), it remains unclear how many Lgr5-positive cells remain in sensory patches of Ysb mutant mice and for how long. It also remains unclear how the repopulation of the flat epithelium should be directed to avoid novel hair cells spreading around the patches, filling in the spiral sulci thus hampering or even disabling the micromechanics of the basilar membrane (Elliott et al., 2018), on which all hearing hinges (Ashmore, 2018).

Since much of the scaffold that is used in current attempts to restore lost hair cells have disappeared in the flat epithelium (Herranen et al., 2020; Pauley et al., 2008; Taylor et al., 2012), we need to start by defining the position of the organ of Corti. A particularly useful cornerstone marker is the position of the inner hair cells adjacent to the habenula perforata (Fig. 2), that allows afferent and efferent nerve fibers to reach inner hair cells, and they need to be positioned near the Hensen’s stripe of the tectorial membrane (Goodyear and Richardson, 2018). Proper positioning inner hair cells are essential for the hydrodynamic flow of endolymph across the tall stereocilia of inner hair cells (Elliott et al., 2018). While reconstitution of the organ of Corti requires regulation of the right cell-types [inner versus outer hair cells (Chessum et al., 2018; Jahan et al., 2018; Wiwatpanit et al., 2018)], the right position (Groves and Fekete, 2017; Jahan et al., 2015a)], and the correct polarity (Elliott et al., 2018; Siletti et al., 2017), we will concentrate here only on the cell positioning issue as a current tractable problem.

4.2. Positioning the organ of Corti along the basilar membrane: a new approach

Although much of the past work has focused on various factors influencing the pattern of the organ of Corti cellular ensemble development (Fritzsch et al., 2006; Groves and Fekete, 2012, 2017), there is almost no information at hand that provides a mechanistic explanation of the positioning of the organ of Corti. Also, the factors guiding the initial regulation of essential transcription factor expression for the organ of Corti development, such as Sox, Pax, Gata, Lmx, Fox, are equally undetermined. And yet, the position of the inner pillar footplate needs to be right on the bony lip of the osseous spiral lamina, and the inner hair cell needs to be at the habenula perforata for normal function (Elliott et al., 2018; Jahan et al., 2018). While future work is needed to define how the organ of Corti is positioned precisely onto the right place during development and what is orchestrating the cascade of genes mediating differentiation, we focus here on genes whose expression remains after the organ of Corti is transformed into a flat epithelium. This strategy allows enhancers of those genes to drive proliferation and differentiation accurately where it is needed. The logic is that the remaining hair cells should not be perturbed during the regeneration attempts.

Work published on several mouse mutants that lose most hair cells, except for small patches, has demonstrated that the replacement of the organ of Corti is by BMP4-positive cells (Dvorakova et al., 2016; Pan et al., 2012; Pan et al., 2011). Importantly, these BMP4-positive cells replace the organ of Corti after Atoh1 deletion has killed hair cells (Chonko et al., 2013) or if hair cells never formed (Dvorakova et al., 2020). In addition, BMP4 shows a gradient of expression with the highest level immediately adjacent to the Henson cells in control and inner sulcus cells after depletion of the organ of Corti (Pan et al., 2012). This gradient could allow using existing BMP4 enhancers, mimicking BMP4 expression (Jumlongras et al., 2012), to drive gene expression proportional to and specific in these cells. In fact, during healthy development, there is a counter-gradient of Fgfs with BMPs, such as Fgf10 medially and Bmp4 laterally, that define the development of distinct cell-types within the organ of Corti (Fritzsch et al., 2006; Groves and Fekete, 2012; Jahan et al., 2015b; Ohyama et al., 2010). Using a viral construct transfection of the BMP4-positive cells that replace the organ of Corti could exploit the BMP4 enhancer to drive Sox2, followed by Atoh1, to initiate pluripotency (Takahashi and Yamanaka, 2006) in the Atoh1 refractory late-stage flat epithelium cells (Kelly et al., 2012). Subsequent downregulation of Sox2 and upregulation of Atoh1 (Dabdoub et al., 2008; Gálvez et al., 2017; Neves et al., 2012) could allow the direct transformation of some flat epithelium cells into hair cells. The BMP4 gradient could also enable the generation of different types of hair cells as this depends on the level of Atoh1 expression as a first step to regulate inner or outer hair cell-specific downstream factors, such as Fgf8 or prestin, respectively (Jahan et al., 2010; Macova et al., 2019). Potentially, the downstream expression of INSM1 (Wiwatpanit et al., 2018) and helios (Chessum et al., 2018) may also depend on Atoh1 expression levels. Of course, identification of a molecular ‘super-regulator’ of the entire multiple steps of the organ of Corti development could greatly facilitate this approach (Ahmed et al., 2012; Li et al., 2020).

Delivering such constructs to humans will have to be done by viral vectors. Currently, most viral vectors used can reliably transfect hair cells with fewer targeting supporting cells (Géléoc and Holt, 2014; Shibata et al., 2016). Moreover, most viral vectors in use for the inner ear have limited cargo capacity that is sufficient to handle single gene loads such as Atoh1 transfection but not multiple genes and their regulatory elements currently in development to ensure proper level of regulation of gene expression, both temporal and spatial (Ausländer et al., 2012; Krawczyk et al., 2020; Sedlmayer et al., 2018; Xie and Fussenegger, 2018). The move forward thus requires at least three steps to be taken:

1. Identify a giant cargo carrying virus able to hold the necessary, multiple genes and their regulatory constructs as well as the BMP4 enhancer to allow targeted, differential regulation of these constructs in a cell to the level of expressing BMP4.

2. Screen viruses for specific docking and cargo delivery to the flat epithelium, sparring the remaining sensory patches that could preserve residual hearing in the apical cochlea currently employed in humans using short cochlear implants (Fattal et al., 2018; Quesnel et al., 2016).

3. Establish that expression of Sox2 followed by Atoh1 is most profound precisely next to the entering nerve fibers and is not excessively counteracted by the BMP4 expression as to disrupt normal hair cell differentiation (Lewis et al., 2018b).

Once these steps are taken, it will be possible to refine the regulatory elements to drive the sophisticated gene regulation needed to differentiate topologically restricted cell types to fill the gaps between remaining patches with new hair cells in mice with the same overall phenotype of the organ of Corti loss at various ages. The reorganization of nerve fibers can help visualize how much regeneration is established (Fig. 2). In addition, proper functional assessment can be initiated in a given mouse before and after restoration to verify how the newly formed hair cells affect frequency-specific auditory brainstem response (ABR) and higher-order information processing.

4.3. Development of viral vectors to achieve organ of Corti restoration from the flat epithelium: a novel future direction

The most common viral vectors currently in use for the development of inner ear gene therapy approaches are adeno-associated virus (AAV) and, less frequently, recombinant adenovirus (rAd). Multiple studies have demonstrated that these vectors are capable of transducing a variety of inner ear cell types and that transduction patterns are dependent on several factors, including viral serotype, route of delivery, and promoter usage [reviewed in (Ahmed et al., 2017; Sacheli et al., 2013)]. Inner hair cells, in particular, appear especially susceptible to transduction by AAV, with some serotypes exhibiting nearly 100% transduction efficiency in vivo (Landegger et al., 2017; Suzuki et al., 2017; Yoshimura et al., 2018). While some studies have been performed in deafened animals, the ability of AAV to transduce dedifferentiated cells of a flat epithelium has not been ascertained.

Most recent studies that demonstrate successful preservation or restoration of inner ear cells or function in deafened models use AAV as the gene delivery vehicle (Budenz et al., 2015; Nist-Lund et al., 2019; Pan et al., 2017; Yoshimura et al., 2019). In each of these studies, AAV was used to deliver a single therapeutic gene or genetic element (miRNA) to cells of the inner ear. While AAV is an ideal candidate for use in some single-gene strategies, especially those targeting inner hair cells, its use in more sophisticated approaches are precluded by a limited capacity for a genetic cargo of only 4.8 kb. Due to this feature, AAV is not a viable option for the delivery of multiple transgenes with sequence elements that will confer cell-type specificity and mediate temporally regulated gene expression. Therefore, the multi-pronged strategy outlined above for the restoration of a properly positioned organ of Corti in Ysb mutant mice would require the development of novel viral vectors.

Helper-dependent adenovirus (HdAd) is structurally identical to adenovirus yet is utterly devoid of all viral coding sequences, solving both primary problems associated with AAV and rAd. The removal of all viral coding sequences leaves capacity for foreign DNA of up to 37 kb, allowing for the delivery of large or multiple transgenes with a single vector. Further, the absence of any viral gene expression renders HdAd non-immunogenic and non-toxic, allowing for long-term transgene expression in the lack of the immune response that is observed with earlier generation adenovirus vectors (Kim et al., 2001; Luebke et al., 2009; Luebke et al., 2001; Muhammad et al., 2012).

While HdAd holds promise for the development of novel vectors for use in the inner ear, significant challenges to the practical implementation of the proposed strategy remain. First, the vector must have the desired tropism, that is, efficient transduction of the dedifferentiated cells of the flat epithelium with minimal transduction of the remaining sensory patches. This will require the identification of a vector serotype that mediates preferential transduction of the flat epithelium through interaction with a cell surface molecule not present on the surrounding sensory cells. Cell-type specificity may also be achieved at the level of gene expression, for example, through the exploitation of sequence elements such as cell-type-specific promoters that drive expression only in the target cell type. These considerations are not trivial given the lack of cell-lines or other systems in which to efficiently screen for tropism in the inner ear. Another significant challenge for the development of a strategy that requires the precise modulation of multiple genes in a temporally (and potentially spatially) controlled fashion is the identification of sequence elements necessary to mediate this regulation. This may require the utilization of different promoters, enhancers, or negative regulators that can be turned on or off to achieve sequential gene expression at times and to the levels necessary to drive reprogramming of the dedifferentiated epithelium.

4.4. Artificial gene regulatory network to transform BMP4 positive flat epithelium cells into hair cells.

As outlined above, simple expression of a single gene or even two genes that are necessary for hair cell differentiation, such as Atoh1 and Pou4f3, will not suffice to drive hair cell differentiation in the adult flat epithelium (Kelly et al., 2012; McGovern et al., 2019; Pauley et al., 2008). In recent years, dramatic progress has been made to define artificial gene regulatory networks (Ausländer et al., 2012; Krawczyk et al., 2020; Sedlmayer et al., 2018; Xie and Fussenegger, 2018). Importantly, some of these networks can simulate the natural oscillation of gene expression, which seems to be so ubiquitous in developing neuronal systems (Kageyama et al., 2018). In addition to oscillation, neuronal differentiation requires a sequential gene activation (Lamb, 2013; Telley et al., 2016) both of which need to be simulated using elements isolated from various life forms (Danino et al., 2010; Ryback et al., 2013; Sedlmayer et al., 2018). Combining all the different elements needed to drive Sox2 and Atoh1 in an oscillatory fashion within BMP4-positive cells of the flat epithelium will require a total cargo load of approximately 25 kb, well within the range of the HdAd viral vectors (37 kb) proposed above.

LuxR = 0.75kb

LuxI = 0.58kb

AiiA = 0.75kb

SOX2 = .95kb

ATOH1 = 1kb

ATOH1 enhancer = 1.3kb

BMP4 enhancer = 16.7kb

GFP = 0.72kb

tdTomato = 1.4kb

CMV promoter = 0.2kb

IRES = 0.6kb

Total ~ 25kb

Given the extra cargo available with the below-outlined construct, the excess capacity could be used to make more complicated fluorescent protein constructs that report, for example, the expression of Sox2 in addition to the Atoh1-EGFP and the virally activated tdTomato to monitor the progression of activation after transfection. Given that targeting BMP4-positive cells to make sure that hair cells differentiate where they need to be on the basilar membrane to be stimulated by the sound invoked movements (Elliott et al., 2018; Reichenbach and Hudspeth, 2014), the regulatory elements driving pluripotency followed by hair cell differentiation need to be at high enough levels to overcome the antagonistic role of BMP4 in hair cell differentiation (Patel et al., 2018; Pujades et al., 2006).

The approach proposed here can be used to first generate a reasonable simulation of an organ of Corti in the neonatal Ysb mutants out of the flat epithelium between the remaining sensory patches (Fig. 2). Using the above-outlined vectors to specifically transfect the ‘flat epithelium’ cells in this mouse model using the BMP4 enhancer (Jumlongras et al., 2012) allows assessment for how far such transformation into hair cells and concomitant organ of Corti-like differentiation in early neonatal animals can go. We propose to employ a well-established synthetic oscillator that allows to dynamically modulate the transcriptional machinery of target genes (Fig. 3). This oscillator is based on the bacterial quorum-sensing system (Sedlmayer et al., 2018).

Fig. 3.

This diagram conceptualizes the repressor/activator artificial regulatory elements needed to drive Sox2 expression Note that the details of the oscillator are regulating the level of Sox2 but in a feedback loop involving several genes and their regulator elements. The depicted synthetic construct provides a way to control the expression of ATOH1 in an oscillatory manner dynamically. This construct consists of LuxR, driven by BMP4 enhancer and LuxR-binding elements in front of acyl homoserine-lactone (AHL) synthase (LuxI), N-acyl homoserine lactonase (Aiia), and SOX2. Together, this leads to basal expression of the elements downstream of LuxR. LuxI synthesizes AHL, which accumulates and determines the LuxR-sensitive threshold. This event results in rapid activation of the LuxR-AHL-mediated transcription of LuxI, AiiA, and SOX2. However, AiiA encodes an enzyme that degrades AHL, thus inactivating the quick-expression caused by AHL. As a result, the system returns to its basal expression and is ready for another oscillation as long as the BMP4-LuxR system remains active. Once Sox2 and Atoh1 have reached a yet to be defined threshold, BMP4 will no longer be active as the transfected cell begins to differentiate into a hair cell. eGFP will serve as a marker for transfection. The construct thus simulates gene expression oscillation as well as sequential gene activation, both of which play essential roles in neurosensory development. Abbreviations: AHL, Acyl-Homoserine Lactone; AiiA, N-acyl homoserine lactonase; BMP4, bone morphogenetic protein 4; LuxI, Acyl-homoserine lactone; LuxR, Transcriptional activator protein LuxR. Modified after (Danino et al., 2010; Krawczyk et al., 2020; Ryback et al., 2013)

In this system, only BMP4 expressing cells will activate via the BMP4 enhancer generation of the LuxR protein that, in turn, will enable the LuxI, AiiA, and Sox2 protein (Fig.3). The proteins generated by LuxI and AiiA will regulate both positive and harmful levels of LuxR protein, initiating an oscillation between high and low levels of LuxR due to different protein decay times (Fig. 3). This oscillation will also drive the level of Sox2 expression, which in turn will activate Atoh1 expression (Alonso et al., 2018; Neves et al., 2012). Atoh1 protein acts as positive feedback on its enhancer (Pan et al., 2012) but also provides a negative feedback loop onto Sox2 (Dabdoub et al., 2008; Kempfle et al., 2016). Note that the construction will be carrying human genes, except for the bacterial genes needed for the oscillator that starts the cascade only in BMP4 expressing cells. Once hair cell differentiation has passed a yet to be defined threshold, the BMP4 expression will be reduced in a negative feedback loop of differentiated hair cells onto BMP4 (Lewis et al., 2018b) and the system will shut down leaving molecular feedback loops in the differentiating hair cells to complete the process.

Once established in its function in neonatal mice carrying the Ysb mutation, the above-outlined system can be expanded, possibly by the addition of extra elements, to function in increasingly older mice. About 12 kb of possible cargo is left in our proposed construct to accommodate the increasing complexity of the artificial gene regulatory network as needed. Once hair cell formation can be induced in all age classes of mice, with future modifications, we envision that this approach is scalable to take advantage of synthetic biology to regenerate the impaired structural architecture of the human ear. To this end, we propose to use human transcription factors to facilitate the transfer of the virally mediated hair cell restoration tested first in mice to humans.

5. Summary and conclusion

We reviewed the impact of various hearing defects and tipped that ARHL, including neurosensory hearing loss, is by far the most prevailing hearing loss, reducing progressively auditory communication in the elderly. Our review of current attempts at combating hearing loss shows dramatic progress but also a focus on less prevalent forms of hearing loss that provide little insight into steps to be taken to restore an old organ of Corti from a flat epithelium that may have replaced the human hearing organ for years. We propose a new approach using the Sox2^Ysb mouse model to serve as a scalable prototype that can allow to first establish in newborn mice how to restore the flat epithelium and sequentially expand this successfully to older stages using newly designed viral vectors. To enable the transfer of mouse model data to humans we propose to use a viral vector approach with human genes that can be targeted to the flat epithelium to deliver the needed large-sized cargo for an artificial gene oscillator to drive first Sox2 expression to prime the flat epithelium to adopt prosensory fate, followed by Atoh1 regulation to initiate hair cell differentiation.