Diagnostic and Therapeutic Applications of Genomic Medicine in Progressive, Late-Onset, Nonsyndromic Sensorineural Hearing Loss

Abstract

The progressive, late-onset, nonsyndromic, sensorineural hearing loss (PNSHL) is the most common cause of sensory impairment globally, with presbycusis affecting greater than a third of individuals over the age of 65. The etiology underlying PNSHL include presbycusis, noise-induced hearing loss, drug ototoxicity, and delayed-onset autosomal dominant hearing loss (AD PNSHL). The objective of this article is to discuss the potential diagnostic and therapeutic applications of genomic medicine in PNSHL. Genomic factors contribute greatly to PNSHL. The heritability of presbycusis ranges from 25 to 75%. Current therapies for PNSHL range from sound amplification to cochlear implantation (CI). PNSHL is an excellent candidate for genomic medicine approaches as it is common, has well-described pathophysiology, has a wide time window for treatment, and is amenable to local gene therapy by currently utilized procedural approaches. AD PNSHL is especially suited to genomic medicine approaches that can disrupt the expression of an aberrant protein product. Gene therapy is emerging as a potential therapeutic strategy for the treatment of PNSHL. Viral gene delivery approaches have demonstrated promising results in human clinical trials for two inherited causes of blindness and are being used for PNSHL in animal models and a human trial. Non-viral gene therapy approaches are useful in situations where a transient biologic effect is needed or for delivery of genome editing reagents (such as CRISPR/Cas9) into the inner ear. Many gene therapy modalities that have proven efficacious in animal trials have potential to delay or prevent PNSHL in humans. The development of new treatment modalities for PNSHL will lead to improved quality of life of many affected individuals and their families.

Introduction

Progressive, late-onset, nonsyndromic sensorineural hearing loss (PNSHL) is the most common sensory deficit and includes several clinical entities such as presbycusis, noise induced hearing loss (NIHL), drug ototoxicity, and inherited forms of hearing loss usually autosomal dominant. Presbycusis, or age-related hearing loss, is a common cause of sensory impairment globally and thus constitutes a major health and socioeconomic burden on all populations (Huddle et al., 2017). Presbycusis affects almost one third of all individuals greater than 65 years of age (Homans et al., 2017). This incidence is increasing as the older subsets of the population increase in size (Wattamwar et al., 2017). Hearing loss in elderly populations has been associated with decreased emotional vitality, mental health issues, higher incidence of dementia, and a more rapid decline in executive function once dementia is present (Gates et al., 2010; Gurgel et al., 2014; Contrera et al., 2016; Su et al., 2017). Other late onset and progressive forms of sensorineural hearing loss, such as noise induced hearing loss (NIHL), are common and constitute a significant burden (Huddle et al., 2017). While most forms of PNSHL have traditionally been considered to be acquired, it is becoming well-understood that genomic factors play a varying (and often substantial) role in predisposition for these conditions (Momi et al., 2015). Presbycusis has been shown to have a heritability ranging from 25% to 75% in twin studies, depending on the type of hearing test used in the study (Momi et al., 2015). AD PNSHL refers to inherited autosomal dominant hearing loss that is progressive, of delayed-onset, and an isolated clinical finding. In addition to contributing to our understanding of the pathophysiology of these forms of hearing loss, understanding genomic factors that contribute to PNSHL has opened the possibility of future predictive and diagnostic testing modalities and new targeted therapies for PNSHL. While most current treatments for PNSHL focus on compensating for hearing loss after hair cell loss, there are no proven preventative or regenerative interventions for high-risk individuals.

Genomic medicine

Genomic medicine is a rapidly evolving field that involves the use of predictive and diagnostic DNA, RNA, or epigenetic testing, as well as related genomically-targeted therapies in the management of pathologies that are not limited to classical Mendelian conditions. Genomic medicine has impacted current clinical areas of medicine, most notably in oncology with the increasing use of somatic genetics to guide targeted therapies across many malignancies (McCarthy et al., 2013).

Genomic medicine has the potential to dramatically improve diagnosis and treatment of neurosensory disorders like deafness and blindness. These disorders share several features that make them ideal for genomic diagnostic and therapeutic approaches. They have well-understood pathophysiology, raising the possibility for targeted therapies. They are common, making them easier to study. They involve anatomic structures that are amenable to local, non-systemic therapies by commonly utilized procedural approaches. Among causes of hearing loss, PNSHL is amenable to preventative and protective interventions because it exhibits delayed onset and is progressive in nature, providing time for intervention between a genetic diagnosis and final symptom manifestation.

Contributions of Genomic Medicine to PNSHL

Genome-wide association studies (GWAS) and other genotyping studies have identified several variants that are associated with presbycusis, including SNPs in the genes TRIOBP (OR=1.132) and SIK3 (total beta = −0.24) (Wolber et al., 2014; Hoffmann et al., 2016). The complexity of applying GWAS results clinically is illustrated by other studies that have found that these variants may exert their effects in a population-specific manner (Luo et al., 2016). A GWAS study conducted in young adult males exposed to occupational impulse noise (firearm training for 10 days) for the first time found that a SNP (rs7598759) in the gene coding for nucleolin (an anti-apoptotic protein) was significantly associated with a change in hearing threshold, with an odds ratio of 12.75 (Grondin et al., 2015). A study identified a SNP in ACYP2 that confers increased risk of cisplatin ototoxicity, with a hazard ratio of 4.5 (Xu et al., 2015). Aminoglycoside ototoxicity is thought to be more common in Chinese populations compared to American populations partly because of differences in the prevalence of mitochondrial DNA A1555G mutation (Liu et al., 2008). Genotyping and other genetic studies have identified polymorphisms in other genes involved with oxidative stress, potassium recycling, and heat-shock proteins that contribute to risk of PNSHL (Liu and Yan, 2007; Sliwinska-Kowalska and Pawelczyk, 2013). Additionally, less deleterious variants in genes that cause inherited forms of hearing loss have been linked to PNSHL (Liu and Yan, 2007). Polymorphisms in genes related to oxidative stress, such as GSTMI, NAT, SOD2 and CAT have been associated with presbycusis and NIHL (Rabinowitz et al., 2002; Fortunato et al., 2004; Ünal et al., 2005; Konings et al., 2007). In patients with presbycusis, variants in the antioxidant enzyme GSTT1 have been shown to correlate with specific high-frequency-steeply sloping audiometric profile (Angeli et al., 2012). Polymorphisms in genes that regulate inner ear potassium homeostasis, of which polymorphisms in KCNE1 and KCNQ4 have been replicated in separate populations, have also been associated with NIHL (Van Laer et al., 2006; Pawelczyk et al., 2009). Interestingly, the ApoE 4/4 genotype has been associated with presbycusis, even when controlling for other ApoE related co-morbidities such as cardiovascular disease, stroke, and cognitive impairment (Kurniawan et al., 2012). This may have mechanistic implications for the relationship between hearing loss and cognitive decline. While genotyping studies have identified possible therapeutic targets by shedding light on important pathophysiologic mechanisms of these forms of hearing loss, the small effect sizes of identified variants make their diagnostic use less feasible. Additionally, the population dependent nature of some associations increases the complexity of diagnostic implementation. Some SNPs contributing to age related hearing loss in populations of European descent have been shown to lose their association in a Han Chinese population (Luo et al., 2016).

Variants in P2X2, a gene encoding an ionotropic purinergic receptor have been implicated in presbycusis and NIHL. P2X2 is expressed in various sites, including the auditory system where it is expressed by outer hair cells (Järlebark et al., 2002). Upregulation of P2X2 decreases outer hair cell signal transduction and has been shown to be protective against noise exposure (Wang et al., 2003; Thorne et al., 2004). This mechanism has been shown to be blunted in older mice thus predisposing to greater hearing loss given equal noise trauma (Telang et al., 2010). In humans, heterozygous loss-of-function mutations in P2X2 (p.V60L) have been associated with AD PNSHL with variable onset in young adulthood or adulthood in several families of variable ethnic origins (Yan et al., 2013; Faletra et al., 2014; Moteki et al., 2015). Yan et al. identified a heterozygous p.V60L missense mutation in P2X2 as causing AD PNSHL in two Chinese families (Yan et al., 2013). Faletra et al. identified a heterozygous p.Gly353Arg variant in an Italian family with a similar phenotype (Faletra et al., 2014). Moteki et al. found an association between P2X2 p.Asp201Tyr and severe sensorineural hearing loss in Japanese patients with MELAS, where patients who lacked the p.Asp201Tyr mutation had normal hearing (Moteki et al., 2015). The more severe phenotype was theorized to be caused by an interaction between the mitochondrial variant causing MELAS causing decreased ATP production and exacerbating the effect of the heterozygous loss-of-function variant in P2X2. No studies exist examining whether this variant may produce a different phenotype in healthy individuals (Moteki et al., 2015). Currently the incidence of such variants in individuals with presbycusis and NIHL is unknown. Many of the patients with P2X2 related AD PNSHL in these reports presented with only mild sensorineural hearing loss, which has the potential to be undiagnosed for long periods of time. Combined with strong heritability data, this raises the possibility that genomic contributions to presbycusis and NIHL may be more significant and more common than previously thought.

Few studies exist employing massively parallel sequencing (MPS) technologies in humans with presbycusis or NIHL. A study screened 32 Korean probands with adult onset, progressive hearing loss that segregated in either sporadic or autosomal recessive fashion using targeted gene panel or whole exome sequencing (WES), and found several patients with variants CDH23 (Kim et al., 2016). Three carried one pathogenic variant described in pre-lingual forms of deafness and one carried a novel rare variant. Patients were affected at a wide range of ages, spanning from 20 to 60 years of age (Kim et al., 2016).

AD PNSHL may be under-recognized as it is unknown to what degree it is misdiagnosed as presbycusis. While over 80 loci have been identified for AD hearing loss, only 38 genes with causative variants have been described (http://HereditaryHearingLoss.org). Compared to AR hearing loss, AD hearing loss tends to be less severe and exhibit more variability between members of an affected family (Liu and Xu, 1994). AD PNSHL is genetically heterogeneous, with identified causative variants coding for proteins with a variety of functions such as connexins (dominant negative GJB2 mutations), extracellular matrix proteins (TECTA and COL11A2), and ion channels (TMC1) (Yan and Liu, 2008). AD PNSHL poses a therapeutic challenge as most cases require disruption of a deleterious protein which largely limits gene therapy approaches to modalities that are gene editing (e.g. CRISPR) or RNA interference (Gurumurthy et al., 2016). However, AD PNSHL is an excellent model for genomic medicine approaches. AD PNSHL has pathophysiology that is analogous to presbycusis, NIHL, and medication ototoxicity in that it is progressive and delayed in onset because of the accumulation of damage to hair cells from an aberrant protein product (Petersen, 2002; Liu and Yan, 2007; Gurumurthy et al., 2016). The progressive nature of AD PNSHL provides for a wide window for detection and intervention, making it an ideal candidate for gene therapy approaches that may then be generalized to other forms of PNSHL (Angeli et al., 2012; Mittal et al., 2017).

Gene expression and epigenetics in PNSHL

Epigenetics refers to the study of regulatory, heritable mechanisms (namely microRNAs, histone modification, and DNA methylation) that affect expression and are independent of changes in the genetic code (Ajonijebu et al., 2017). Due to the known role of environmental factors in PNSHL, epigenetic mechanisms may have important contributions (Van Eyken et al., 2007). Expression studies have identified numerous genes with roles in immunity and apoptosis that are differentially expressed in the peripheral blood of patients with presbycusis (Dong et al., 2014). MicroRNAs (miRNA) are small non-coding RNA molecules consisting of about 22 nucleotides that are responsible for silencing mRNA expression and have roles in development and survival of inner ear hair cells (Friedman and Avraham, 2009; Zhao et al., 2017; Mittal et al., 2019). In one report, a Spanish family with AD PNSHL was found to have a point mutation in MIR96 that was absent in controls (Mencia et al., 2009). MIR96 is thought to modulate the expression of many genes in the inner ear through various mechanisms including direct silencing of mRNA targets and indirectly through changes in expression of genes that do not code for an miR96 target by mechanisms such as histone modifications (Friedman and Avraham, 2009). Mouse studies of presbycusis have identified several differentially expressed miRNAs in the cochlea, where many of the upregulated miRNAs are involved in pro-apoptotic pathways and downregulated miRNAs tend to be involved in proliferation and differentiation pathways (Zhang et al., 2013). Plasma levels of miR-185–5p and miR-451a have been found to be downregulated in individuals with NIHL (Ding et al., 2016). Conversely, another study found that mmu-miR-130b-3p, mmu-miR-27a-3p, mmu-miR-27b-3p and mmu-miR-721 downregulate PPARG leading to the production of oxygen radicals and eventual cell death in a mouse model of PNSHL; the authors hypothesized that inhibition of those miRNAs could be a therapeutic strategy in this form of hearing loss (Zhang et al., 2019).

DNA methylation refers to the addition and removal of methyl groups to CpG islands of DNA via methylase and demethylase enzymes. Addition of methyl groups blocks transcription factor binding and silences gene expression (Friedman and Avraham, 2009). In histone modification, enzymes instead manipulate the histone proteins through mechanisms such as acetylation and deacetylation, which is controlled by coordinating actions of histone deacetylase enzymes (HDACs) and histone acetyl transferase enzymes (HATs) (Janssen et al., 2017). Knowledge of the role that these mechanisms play in inner ear biology could lead to identification of therapeutic targets. In one study, DBA/2J mice (model of PNSHL) were received valproic acid (HDAC inhibitor) and methionine (methyl donor) and showed improved auditory brainstem response (ABR) thresholds compared to controls (Mutai et al., 2015). Microarray revealed differential expression of 299 genes following injection, including increased expression of a zinc importer gene, Zip4, that plays a role in signaling in various parts of the mouse inner ear (Mutai et al., 2015). Upregulation of Zip4 by (−)-epigallocatechin-3-O-gallate caused similar improvement in ABR thresholds as a result of these epigenetic modifications (Mutai et al., 2015). Oxygen radical damage is a major contributor to PNSHL: another study has shown that increased methylation of superoxide dismutase 2 (SOD2) led to mitochondrial damage and cell death in a rat primary cell culture model of inner-ear marginal cells. A methylation inhibitor (5-azacytidine) was able to prevent this damage (Li et al., 2020).

Diagnostic applications of genomic medicine in presbycusis

A variety of genetic diagnostic modalities are routinely employed for cases of inherited, congenital, or early onset sensorineural hearing loss (Table 1). Microarrays are often used as first-line testing in genetically heterogeneous conditions such as deafness (Yan et al., 2017). Their major advantage is that they allow for affordable, rapid, simple testing for multiple variants simultaneously with less bioinformatic requirements than MPS approaches (Yan et al., 2017). However, the advent and steady improvement of MPS technologies (with progressively decreasing costs) has opened the door to new options. In hearing loss suspected to have a genetic component, an MPS panel including exonic regions for all known deafness genes reached a diagnostic rate of 42%, but this decreased to 22% in cases of adult onset hearing loss (Shearer et al., 2013). Still, gene panels currently are the first-line diagnostic tool. Compared to WES and WGS, gene panels have better coverage over targeted regions, lower cost, decreased bioinformatic requirements, and a lower risk of incidental findings (Sloan-Heggen et al., 2016). Issues with gene panels are largely related to significant variability among tests. Panels often vary by which genes are included and whether they detect copy number variants (CNVs) (Table 1).

Table 1.

Pros and cons of various diagnostic modalities currently available for PNSHL.

Diagnostic Modality	Time and Cost	Pros	Cons
Gene Panel	Weeks ~$500	High sensitivity for genes in panel; Sampling of most common genes for lower price	Genes not on the panel are not screened; Copy number and structural variants not always analyzed
Whole Exome Sequencing	Weeks-months ~$1000	Analyze variants in coding regions	Can miss intron variations; Higher cost may be prohibitive
Whole Genome Sequencing	Weeks-Months $1000	Analyze variants in entire genome	Complex computational tools needed for analysis; Ethical issues of incidental findings

Decreasing costs of MPS are increasing the feasibility of WES as a first-line option for genetic testing in hearing loss. Some reports of the use of WES in non-syndromic hearing loss suggest that WES has higher diagnostic rates than targeted gene panels. Bademci et al. found causal variants in 56% of patients from 160 multiplex families with AR NSHL and previous negative GJB2 testing (Bademci et al., 2016). Of these, 54% of variants were novel. WES has several benefits over gene panel such as may have a higher diagnostic rate and can detect variants in regions of the exome not covered by a panel. In addition, statistical tools now exist to improve the ability of WES to detect CNVs, and WES has been used to successfully detect pathogenic CNVs in patients with congenital hearing loss (Fromer et al., 2012; Bademci et al., 2014). Whole genome sequencing (WGS) has been used in a research setting to screen for novel candidate genes and variants as well as to provide new insights into the molecular characterization of age-related hearing loss (Vuckovic et al., 2018; Sineni et al., 2019) but has not been implemented in a clinical setting in patients with PNSHL. While the ability of WGS to detect mutations in noncoding portions of the genome (Table 1) may be of use eventually, currently this may be undesirable as these variants are often poorly understood. However, studies comparing its diagnostic ability to that of WES in other genetically heterogeneous conditions such as intellectual disability have shown it to have a superior diagnostic rate with improved detection of CNVs (Gilissen et al., 2014).

Clinical and personal utility, burden, cost-effectiveness, and disparities

Given that these technologies have yet to be implemented in a systematic fashion for most forms of PNSHL, there are no studies assessing to what extent they will influence decisions about follow-up, further testing or treatment. While genetic diagnosis has been shown to play a role in prognosis in cochlear implantation in some congenital forms of inherited sensorineural hearing loss, this has not been determined in individuals with PNSHL and positive genetic testing (Vivero et al., 2010). It is conceivable that knowledge of increased susceptibility to PNSHL could influence patient management in other ways such as increasing follow-up frequency, decreasing the threshold for auditory testing, decreasing time from onset of hearing loss to treatment, and creating stricter occupational noise protection requirements or recommendations for at-risk individuals. By identifying an etiology for hearing loss, it also has the potential to reduce the need for additional testing and imaging in some patients. It may hold additional personal utility for some patients who are interested in prognostic information as to the expected progression and severity of their hearing loss (Arnos, 2008). This must be balanced against the possibility of increased costs, psychological impacts of genetic testing, the possibility of some patients being tested unnecessarily and, with the increasing use of MPS technologies, incidental findings or variants of unknown significance (Arnos, 2008). Until the true prevalence of clinically actionable genetic diagnoses in individuals with PNSHL is known, it will be difficult to balance these factors and develop standardized testing guidelines.

Cost-effectiveness is highly dependent on the specific testing modality employed and on the diagnostic yield based on clinical context. For example, gene panels have been shown to have much lower diagnostic yield in cases of asymmetric hearing loss thought to have an inherited component when compared to symmetric hearing loss (Shearer et al., 2013). Panels have also been shown to have lower diagnostic rates in hearing loss that is not congenital, which includes PNSHL. For example, diagnostic rates are around 28% in cases of genetic childhood- or adult-onset hearing loss, compared to more than 40% in all cases (Sloan-Heggen et al., 2016). Panels may sometimes be useful in situations where previous testing is negative. In one study, a gene panel provided a diagnosis in 13% of patients with PNSHL thought to have a significant genetic component but negative single-gene testing for common mutations, sometimes through detection of CNVs (Shearer et al., 2014; Gu et al., 2015). Despite this added benefit, panels have comparable costs to traditional single-gene testing (Shearer et al., 2014; Moteki et al., 2015). Compared to other MPS technologies, gene panels have greatly lower costs (may be less than US$500 compared to about than US$1000 for WES and more for WGS in clinical laboratories) and lower turnaround times (Tekin et al., 2016; McDermott et al., 2019).

Traditional genetic testing algorithms for many genetic conditions have often been based on results of studies in individuals of European descent. However, it is well known that variants that are common in highly studied populations may be less common in other less-studied populations (Samanich et al., 2007; Liu et al., 2008). For example, variants in GJB2 are much less common in African American and Caribbean Hispanic patients with genetic hearing loss (Samanich et al., 2007). Additionally, some variants have been associated with presbycusis in certain populations but found to be unrelated in others (Luo et al., 2016). These factors naturally result in the potential for differences in diagnostic rates and testing burden across ethnic groups. Microarrays may avoid this to some extent when designed to detect common mutations across various ethnicities (Yan et al., 2017). MPS techniques eliminate the need to design gene-by-gene testing algorithms and decrease testing disparities (Shearer and Smith, 2015). However, these techniques have still been shown to vary in efficacy by ethnicity. One gene panel demonstrated major differences in diagnostic rates by ethnicity, with greatest efficacy in patients of Asian and Middle Eastern descent, followed by Caucasian descent, and with significantly lower diagnostic rates in African American patients (Sloan-Heggen et al., 2016).

Therapeutic genomic medicine applications in PNSHL

Genomic medicine and current treatments for PNSHL

Current therapeutic applications of genomic medicine in PNSHL are limited. In some cases, a genetic diagnosis may help predict the need for or success of an established therapeutic modality such as cochlear implantation (Vivero et al., 2010; Eshraghi et al., 2012). Genomic medicine approaches have the potential to be used in combination with established therapeutic modalities with possible additive or synergistic effects. For example, a known mechanism for decreased cochlear implant performance is auditory neuron degeneration in the face of loss hair cells and subsequent decrease in neurotrophies that they normally release (Wise et al., 2010; Fukui et al., 2012). Studies of BDNF gene therapy in a mouse model of genetic deafness have shown preservation of auditory neurons in the absence of hair cells and sprouting in these neurons, making it a strong candidate for improvement of cochlear implant performance (Fukui et al., 2012). Similarly, cochlear BDNF or neurotrophin-3 gene therapy in guinea pigs with aminoglycoside induced deafness increased auditory neuron survival (Wise et al., 2010). Cochlear implantation itself can be used as for gene delivery through direct electroporation of hair cells and has been shown to improve hearing outcomes determined by ABR in guinea pigs exposed to local aminoglycoside (Eshraghi et al., 2012; Pinyon et al., 2014).

Gene therapy modalities and therapeutic routes for PNSHL

Gene therapy refers to the therapeutic use of nucleic acids such as DNA or RNA in human disease. Gene therapy for hearing loss requires delivery of nucleic acids (naked or with a vector) into the inner ear (Table 2). These therapies can be categorized based on whether they edit the target cells’ genome (e.g. correct a point mutation) or supply exogenous nucleic acid sequences for some other purpose. Of the genome editing modalities, CRISPR/Cas9 is most likely to find therapeutic application in humans. Its advantages include greatly decreased time, labor, and cost requirements compared to previous genome editing techniques (Zou et al., 2015). CRISPR/Cas9 can be used to both counteract the gain-of-function effect of a dominant mutation or repair a recessive mutation and rescue normal protein expression or function. A major concern with genome editing techniques is the potential for off-target genome editing effects. Another benefit of CRISPR gene editing is that specificity has been progressively optimized to the point where off-target effects are uncommon (Gurumurthy et al., 2016; Kleinstiver et al., 2016; Slaymaker et al., 2016; Mittal et al., 2017). As a proof-of-concept, direct injection of CRISPR/Cas9 reagents into one-cell mouse embryos has been used successfully to correct a mutation in Cdh23 that predisposes mice to age-related hearing loss with reversal of phenotype and without detection of any off-target effects by WGS (Mianné et al., 2016).

Table 2.

Comparison of therapeutic modalities for PNSHL.

	Viral Delivery			Non-Viral Delivery
	DNA Delivery	RNAi	Differentiation	Naked Nucleic Acids	Electroporation	Lipofection
Mechanism	Direct delivery of deficient gene	Knockdown gain-of-function mutations	Gene to promote HC generation	DNA or RNA trans-tympanic injection	Cochlear implant DNA electroporation	Liposome delivery of DNA or RNA
Difficulties	Limited to ~5kb; Mutagenesis	Repeat dosing	Low efficiency	Low efficiency; Repeat dosing	Cochlear implant needed	Not specific to target cell
Example	VGLUT3 delivery	Tmc1 knockdown	ATOH1 delivery	siRNA knockdown of NOX3	BDNF/neurotropin after CI stimulation	siRNA knockdown of GJB2
Testing Phase	In vivo mouse	In vivo mouse	Clinical Trial	In vivo mouse	In vivo guinea pig	In vivo mouse

Non-genome editing modalities include delivery of exogenous genetic material for the purpose of expressing a deficient protein (e.g. treatment of a loss-of-function mutation) or therapeutic molecule (e.g. a neurotrophic factor) (Sacheli et al., 2013). RNA interference (RNAi) is another non-genome editing modality that may be of use in the inner ear for knocking-down gene expression (e.g. gain-of-function mutations) (Mukherjea et al., 2010). In many clinical situations, CRISPR/Cas9 may be advantageous compared to non-genome editing techniques, as it can have a permanent therapeutic effect without being expressed permanently (Zou et al., 2015). In contrast, exogenous expression of a gene and RNAi require long-term expression for a permanent biologic effect, and thus must utilize a viral expression system that can integrate into the genome. When a transient effect is required, such as for temporary treatment after an ototoxic insult or temporary expression of a transcription factor, non-genome editing technologies may be useful (Izumikawa et al., 2005; Mukherjea et al., 2010).

Gene therapy for SNHL is more feasibly accomplished by local rather than systemic routes, as the cochlea is accessible to local pharmacotherapy through several approaches (Gurumurthy et al., 2016). Systemic therapies must cross the blood-cochlea barrier in order to exert their action (Sacheli et al., 2013). Several local therapeutic approaches exist for access to the cochlea, including intratympanic injection with diffusion across the round window membrane, round window membrane injection, stapedotomy with oval window injection, and cochleostomy with injection (Sacheli et al., 2013; Dai et al., 2017). Oval window and round window membrane injections have been shown to be safe in rodent and rhesus monkey models, and the stapedotomy with oval window injection approach is currently being employed in the human trial NCT02132130 (Dai et al. 2017; https://clinicaltrials.gov/ct2/show/NCT02132130; Klickstein L. 2013).

Therapeutic viral gene delivery for PNSHL

Adeno-associated virus (AAV) is a commonly used viral vector. It is not pathogenic in humans and has the potential to incorporate into the host cell genome allowing for durable transgene expression but creating a risk for disrupting the native sequence. AAV has been used successfully in humans in the comparable setting of local gene therapy for vision loss (Maguire et al., 2008; Testa et al., 2013; MacLaren et al., 2014; Edwards et al., 2016). These studies have evidenced that viral gene therapy can be safe and effective in humans. For example, AAV vectors have been used to treat Leber congenital amaurosis safely and with improvements in vision that have been sustained throughout three years of follow-up (Maguire et al., 2008; Testa et al., 2013). In human trials of AAV gene therapy for choroideremia, one third of patients showed immediate improvement that was stable at 3.5 years follow-up (MacLaren et al., 2014; Edwards et al., 2016). The cochlea and retina are similar in that they are amenable to local gene therapy by minimally invasive approaches (Table 1).

A limitation of AAV is that it can only deliver less than 5 kb of DNA (Chien et al., 2015). Other viral vectors have not been favored because of high immunogenicity (herpes simplex and vaccinia) or requirement of mitotically active host cells (lentivirus) and poor genome site integration specificity (lentivirus) (Kesser et al., 2008; Sacheli et al., 2013; Chien et al., 2015). Several AAV have been used to target specific populations of neonatal and adult mammalian inner ear cell types in in vivo animal models (Shu et al., 2016). This raises the potential for the selection of vectors depending on the genetic cause of hearing loss and the cellular population where the affected protein is specifically expressed (Shu et al., 2016). Transduction efficiency and vector immunogenicity (although less of an issue than for some other viruses) have been limitations of conventional AAV vectors. A synthetic AAV vector, Anc80L65, exhibits more efficient transduction into mouse inner and outer hair cells while preserving hearing and being antigenically distinct when compared to conventional AAV vectors (Landegger et al., 2017).

Viral gene therapy has been used in various hearing loss models successfully (Table 2). For example, Yu Q et al. showed that Connexin26 gene therapy in GJB2 knockout mice restores the cochlear gap junction network and reduces cell death histologically but with no detectable improvements in ABR response (Yu et al., 2014). Askew et al. treated mice deafened by loss-of-function (analogous to AR DFNB 7/11) and gain-of-function (analogous to PNSHL DFNA 36) mutations in TMC1 with AAV-mediated TMC2 gene therapy (Askew et al., 2015). TMC2 is only expressed developmentally and has a similar function to TMC1. Functional TMC2 improved ABRs in mice with deleted and aberrant forms of TMC1 (Askew et al., 2015). In VGLUT3 knockout mice, cochlear delivery of an AAV vector with the VGLUT3 transgene restored ABR thresholds and startle response (Akil et al., 2012). This model is only partly analogous to DFNA25 (PNSHL) which is caused by AD, missense mutations in VGLUT3 rather than the null mutation in these mice (Akil et al., 2012). The Anc80L65 synthetic AAV vector has been used to successfully restore hearing and balance in a mouse model of Usher syndrome by delivery of wildtype Ush1c with greater than 80% transduction efficiency (Pan et al., 2017).

Viral vectors have also been used for cochlear RNAi through miRNA injection to prevent hearing loss in the Beethoven mouse, a model of AD PNSHL caused by gain-of-function mutations in Tmc1 (Table 2) (Shibata et al., 2016). This mutation is orthologous to the variant that causes DFNA36 in humans, which may have childhood or adult-onset hearing loss that is progressive and severe by the age of 50 (Shibata et al., 2016). A single intracochlear injection was sufficient to significantly attenuate hearing loss by down-regulating the expression of the dominant allele (Yoshimura et al., 2019). RNAi is thus a promising approach for the treatment of hearing loss caused by gain-of-function mutations, as is the case for most AD PNSHL.

Viral gene therapy has potential for common forms of PNSHL such as presbycusis, NIHL, and medication ototoxicity where the therapeutic aim is regeneration of hair cells and not the correction of a single pre-existing genetic defect (Table 2). For example, ATOH1 (also known as HATH1) is a transcription factor important in human hair cell differentiation and has shown therapeutic benefit in mammalian models of NIHL and ototoxicity (Izumikawa et al., 2005; Yang et al., 2012; Richardson and Atkinson, 2015). Izumikawa et al. showed that ATOH1 gene therapy with an AAV vector improved hearing by ABR in mature mice deafened by ototoxic medications by causing trans differentiation of nonsensory cells into hair cells (Izumikawa et al., 2005). Yang et al. showed significant improvement in ABR thresholds with round window injections of ATOH1 with an AAV vector one week after noise exposure in a guinea pig model of NIHL (Yang et al., 2012). Some studies, however, have yielded conflicting results. Atkinson et al. used AAV vectors to deliver ATOH1 alone or in combination with BDNF or neurotrophin-3 through a cochleostomy in guinea pigs deafened by a combination of loop diuretic and aminoglycoside (Atkinson et al., 2014). While they found that treated inner ear cells expressed markers of hair cell differentiation, there was no decrease in ABR thresholds at 3 weeks post-treatment (Atkinson et al., 2014). It is difficult to compare these studies as the disease models may represent different degrees of hair cell death. For example, in Izumikawa et al., baseline ABR thresholds in noise-exposed mice were not as severely elevated as in Atkinson et al. after ototoxic exposure (Izumikawa et al., 2005; Atkinson et al., 2014). This may suggest that ATOH1 gene therapy may be feasible depending on the severity of hair cell loss (Atkinson et al., 2014; Richardson and Atkinson, 2015).

Currently, there is an ongoing phase 1 and 2 clinical trial, NCT02132130, examining the use of ATOH1 (Hath1) gene therapy in adults with bilateral severe to profound SNHL without a specific genetic etiology (https://clinicaltrials.gov/ct2/show/NCT02132130)(Klickstein, 2013). While many of the ATOH1 gene therapy animal studies were conducted with the use of AAV vectors, the human trial is being carried out with the adenovirus vector Ad5. Ad5 does not have the same 5 kb size restriction that AAV has. It also does not integrate into the genome at all, so there is no risk of insertional mutagenesis, which has been one of the greatest concerns preventing human gene therapy trials from progressing (Praetorius et al., 2009; Staecker et al., 2011). Permanent ATOH1 expression is not needed in this therapeutic paradigm, as short-term expression has been sufficient to induce transdifferentiation into hair cells and restore hearing in the studies where it has worked (Izumikawa et al., 2005; Yang et al., 2012). Lower volumes are required when using Ad5, as it has better transduction efficiency and thus carries less risk of cochlear damage during intra-labyrinthine injection (Praetorius et al., 2009; Staecker et al., 2011). While adenovectors are believed to be more immunogenic than AAV vectors, the low dose requirement and modifications to the virus are thought to counteract this (Praetorius et al., 2009; Staecker et al., 2011). NCT02132130 will involve an intra-labyrinthine injection of ATOH1 delivered by Ad5 at varying doses. ATOH1 will be expressed under control of the glial fibrillary acidic protein (GFAP) promoter, which should limit expression to inner ear support cells expected to undergo transdifferentiation into hair cells in response to therapy (Klickstein, 2013). If shown to be safe, it may open the door for other inner ear viral gene therapy trials (Table 2).

Non-viral genomic therapeutics for PNSHL

Several other gene-delivery techniques have been applied to the inner ear (Table 2). The accessibility of the inner ear to local therapy allows for other gene delivery modalities that would not be feasible for systemic gene delivery, such as direct application of naked nucleic acids and electroporation. There are few reports of successful direct application of naked nucleic acids to the inner ear. A study performed trans-tympanic injections of siRNA targeting expression of NOX3 prior to administration of systemic cisplatin therapy to mice (Mukherjea et al., 2010). NOX3 is responsible for the majority of reactive oxygen species production in the inner ear (Mukherjea et al., 2010; Kaur et al., 2016). The transient effect of RNA interference therapy two days prior to cisplatin was sufficient to prevent cisplatin-induced expression of NOX3 and associated apoptotic markers, decrease damage to outer hair cell structure on electron microscopy, and reduce post-cisplatin ABR threshold shift in a dose dependent manner (Mukherjea et al., 2010). RNAi can be accomplished transiently or constitutively depending on whether the miRNA or siRNA is delivered directly into the cochlea, or through the use of a viral vector (Mukherjea et al., 2010; Shibata et al., 2016). RNAi thus has the potential to be used preventatively in high risk patients, such as patients expected to experience progression of NIHL in unavoidable high noise settings (e.g. military) or patients undergoing chemotherapy or treatment with aminoglycosides that have already experienced some ototoxicity (Rybak et al., 2012). The transient and local nature of this therapy is important in many cases, as systemic inhibition of NOX3 expression could interfere with the effectiveness of chemotherapy or aminoglycoside therapy (Rybak et al., 2012). Transtympanic injection is generally safe and well-tolerated, and patients who may require multiple treatments could feasibly have pressure equilibration tubes placed for the duration of their treatment, creating easy access to the middle ear space with liquid drops rather than injections. Studies of RNAi prevention of ototoxicity have largely focused on pretreatment and it is unclear to what extent this approach can be used to prevent or reverse hearing loss after ototoxic exposure (Mukherjea et al., 2010; Rybak et al., 2012).

Cochlear electroporation refers to the delivery of genetic material into inner ear cells through the use of a current that temporarily increases the permeability of inner ear cell membranes. Pinyon et al. demonstrated that a cochlear implant could be used for electroporation and transduced the cochleae of guinea pigs deafened by ototoxic agents with a naked DNA construct coding for BDNF (Pinyon et al., 2014). This stimulated growth of spiral ganglion neurites in the direction of the cochlear implant electrodes, improving cochlear implant performance on ABR (Pinyon et al., 2014). More recent studies have demonstrated that direct delivery of BDNF and neurotrophin to the round window after a prolonged 95 dB sound stimulus leads to improved preservation of ABR (Sly et al., 2016). The use of gene therapy for the continuous endogenous production of neurotrophic factors may be protective in environmental and genetic causes of progressive hearing loss with a broad range of etiologies and may improve prosthesis performance when combined with a cochlear implant.

Lipid-mediated delivery of genetic material is also a possible alternative to direct viral delivery for local inner ear gene therapy (Table 2). Compared to direct delivery of nucleic acids, liposomes protect oligonucleotides from nucleases and increase their permeability (Dass, 2002). Cationic lipocomplexes can be used to deliver genetic material, proteins, and viral vectors by diffusion through the round window membrane without violating the inner ear surgically. While neither approach creates a significant inflammatory response, application of lipocomplexes to the round window membrane is less traumatic to the inner ear when compared to direct viral injection (Maeda et al., 2009). This approach also leads to less transgene expression and preferential editing in outer hair cells at the basal turn of the cochlea (Zuris et al., 2015). While this may limit the applicability of this approach to some forms of hearing loss, the basal turn of the cochlea is primarily affected in most forms of PNSHL. Zuris et al. used cationic lipid transfection reagents to deliver Cre recombinase (fused to anionic proteins) and Cas9-guide RNA complexes into the mouse inner ear in vivo. This achieved a gene modification rate of 20% in vivo. Compared to transfection of a plasmid, the action of Cas9-guide RNA complexes was transient, leading to better genome editing site specificity (Zuris et al., 2015). Lipocomplexes result in only a transient effect of the delivered genetic material and thus a transient clinical effect, unless a genome editing system such as CRISPR/Cas9 is employed, in which clinical effect should last the lifespan of the edited cells and all descendant cells (Chien et al., 2015). This is a desirable trait for most potential therapeutic applications of the CRISPR/Cas9 system. While viral delivery can be used when continuous effect of Cas9-RNA is warranted (such as when the goal is continuous correction of mRNA), a continuous effect is not necessary in most other cases, as it would not confer any additional therapeutic benefit after the desired change has been made to the genome and would potentially increase the risk for off-target effects (Zou et al., 2015). This approach has been used therapeutically in a mouse model of AD PNSHL. Gao et al. used cationic lipid delivery of the CRISPR-Cas9 system to disrupt mutant TMC1 in the Beethoven mouse and prevent hearing loss. This approach was selective for the mutant allele of TMC1, despite a difference of only one base pair from the normal allele (Gao et al., 2018). CRISPR-Cas9 may thus be a valuable alternative to RNAi when permanent disruption of a mutant allele is warranted.

Cationic lipocomplexes have also been used to deliver a plasmid coding for a mutant allele of GJB2 (causing hearing loss through a dominant negative effect) and siRNAs designed to downregulate the expression of this allele into the inner ear in vivo (Maeda et al., 2005). In this model, RNA interference (RNAi) through this approach achieved 70% decreased expression of mutant GJB2 without having any effect on expression of endogenous wild type GJB2 (Maeda et al., 2005). This decrease in expression was sufficient to prevent hearing loss. RNAi may thus be achieved without violation of the inner ear space through the use of lipocomplexes, making it a promising approach for hearing loss where gain-of-function mutations have a role (Maeda et al., 2005). The duration of gene expression or RNAi by lipocomplex delivery of plasmids encoding siRNAs is shorter than if this approach was carried out using a viral vector, but greater than that of siRNAs injected directly into the middle ear (Maeda et al., 2005; Mukherjea et al., 2010; Shibata et al., 2016). Thus, the delivery technique for RNAi may be varied depending on the specific clinical entity being treated. For example, in acute hearing loss after noise trauma or exposure to an ototoxic compound, transient interventions have been shown to be effective in preventing hearing loss (Mukherjea et al., 2010; Mizutari et al., 2013). Lipocomplex RNAi may be ideal for situations where an ototoxic insult is expected to be ongoing for some additional time but not permanent.

Conclusions and Future Directions

While there are currently few applications of genomic testing for diagnosis of most forms of PNSHL, many potentially relevant variants have been identified. These variants may have diagnostic utility once targeted preventative or therapeutic interventions are developed for PNSHL. AD PNSHL is an excellent model for developing therapies for presbycusis, NIHL, drug ototoxicity and inherited forms of hearing loss. Current early human trials may open the door for determining the efficacy of viral and non-viral gene therapy modalities that have shown great promise in animal models. Genomic medicine is poised to rapidly gain clinical applicability in treating PNSHL by allowing for early diagnosis and timely intervention to prevent or delay the onset of hearing loss, thereby reducing the global burden of adult-onset progressive hearing loss.

Research Highlights.

• Progressive, late-onset, nonsyndromic, sensorineural hearing loss (PNSHL) is the most common cause of sensory impairment globally.

• Discuss the potential diagnostic and therapeutic applications of genomic medicine in PNSHL.

• Viral and non-viral gene delivery approaches are being used for PNSHL in animal models and clinical trials.

• Development of new treatment modalities for PNSHL will lead to improved quality of life of many affected individuals and their families.

Abbreviations

AAV

Adeno-associated virus

CI

Cochlear Implant

CNVs

copy number variants

GWAS

Genome-wide association studies

miRNA

microRNAs

NIHL

Noise induced hearing loss

PNSHL

Progressive, late-onset, nonsyndromic, sensorineural hearing loss