Genetics of hearing loss: Allelism and modifier genes produce a phenotypic continuum



Recent genetic and genomic studies have greatly advanced our knowledge of the structure and function of genes involved in hearing loss. We are starting to recognize, however, that many of these genes do not appear to follow traditional Mendelian expression patterns and are subject to the effects of allelism and modifier genes. This review presents two genes illustrative of this concept that have varied expression pattern such that they may produce either syndromic or nonsyndromic hearing loss. One of these genes, cadherin 23, produces a spectrum of phenotypic traits, including presbycusis, nonsyndromic prelingual hearing loss (DFNB12), and syndromic hearing loss as part of Usher syndrome (Usher 1D). Missense mutations in CDH23 have been associated with presbycusis and DFNB12, whereas null alleles cause the majority of Usher 1D. Modifier gene products that interact with cadherin 23 also affect the phenotypic spectrum. Similarly, allelsim in the gene encoding wolframin (WFS1) causes either a nonsyndromic dominant low-frequency hearing loss (DFNA6/14/38) or Wolfram syndrome. Missense mutations within a defined region are associated with DFNA6/14/38, while more severe mutations spanning WFS1 are found in Wolfram syndrome patients. The phenotypic spectrum of Wolfram syndrome is also hypothesized to be influenced by modifier genes products. These studies provide increasing evidence for the importance of modifier genes in elucidating the functional pathways of primary hearing loss genes. Characterizing modifier genes may result in better treatment options for patients with hearing loss and define new diagnostic and therapeutic targets. Anat Rec Part A, 2006. © 2006 Wiley-Liss, Inc.

Severe hearing impairment affects 1 in 1,000 newborns and a further 1 in 1,000 children suffer hearing impairment significant enough to affect speech and language development (Fortnum et al.,2001). Approximately 50% of prelingual hearing loss is attributed to hereditary factors. Hereditary hearing loss is divided by clinical phenotype into syndromic, in which hearing loss is concurrent with malformation of other organ systems, and nonsyndromic, in which hearing loss is the only disease phenotype. Syndromic and nonsyndromic hearing losses are classified by modes of inheritance including autosomal dominant, autosomal recessive, X-linked, and mitochondrial or maternally inherited. Nonsyndromic forms of hearing loss have systematic names based on mode of inheritance including DFNA for autosomal dominant, DFNB for autosomal recessive, and DFN for both X-linked and mitochondrial. To date, over 80 genes have been identified in hearing loss that divide almost evenly between syndromic and nonsyndromic forms (Van Camp et al.,

The study of hereditary hearing loss is difficult due to the structural and functional complexity within the hearing apparatus (Friedman et al.,2000). In particular, sensorineural hearing loss is the common heterogeneic result of a multitude of gene mutations that affect inner ear structures (for definitions of genetic terms, see Table 1). Heterogeneity also affects subsets of hearing loss such as Usher syndrome type 1, in which seven different genes or loci have been identified as producing the same syndromic phenotype (Adato et al.,2005). In contrast, mutations in one gene may produce multiple phenotypes, which is known as allelism. Allelism affects many of the identified hearing loss genes. Some hearing loss-related genes display extreme allelism through association with both syndromic and nonsyndromic phenotypes (Table 2). Additionally, the same allele may produce different hearing loss characteristics in different families. This attests to the pleiotropy of a specific allele and is commonly due to the action of modifier genes.

Table 1. Genetic & genomic terminology
AlleleDifferent DNA sequences in a gene or locus.
AllelismDiffering phenotypes resulting from different alleles of the same gene. Also known as allelic heterogeneity.
Compound HeterozygoteCompound heterozygotes for recessive disease alleles usually produce the recessive disease phenotype. Different heterozygous alleles occurring within the same gene or locus.
CongenicAn animal model in which a locus from one strain has replaced the homologous locus of a different strain.
ConsomicAn animal model in which one complete chromosome from a donor strain has replaced the homologous chromosome of a host strain.
DigenicTwo heterozygous alleles of differing genes acting together to produce a phenotype. The alleles are usually recessive.
Dominance modificationChanges in the association between the dominant or recessive genotype and the severity of the disease phenotype.
Dominant AlleleAn allele that creates a phenotype regardless of the matching allele.
EpistasisThe allele of one gene masks the phenotype of a second gene.
ExpressivityVariability in disease severity for allele-carrying targets.
F1 hybridThe first filial generation produced by crossing two parental strains.
F2 hybridThe progeny of F1 hybrid crosses.
GeneA segment of a chromosome that produces a protein. Human genes are displayed in italics with all upper case letters. Mouse genes are displayed in italics and only begin with an upper case letter.
HeterogeneityIndependent mutations in two or more genes that produce the same phenotype. Also known as locus heterogeneity.
IsoformA version of a protein different from others due to mutational or normal alterations (e.g. transcriptional or translational processing).
Linkage AnalysisLocating a locus by measuring recombination rates between phenotypic and genetic markers.
LocusA segment of DNA containing not more than a few genes. A locus is generally denoted by genetic markers and the size of DNA between these markers.
Missense MutationA mutation resulting in a single amino acid change in the gene product.
Nonsense MutationA mutation causing the creation or removal of an mRNA stop codon.
NullDesignates a gene allele containing one or more mutations that result in the complete lack of protein expression.
OrthologousA gene or locus in one species that provides the same function in a different species.
PenetranceThe proportion of allele-carrying targets that display the disease phenotype. If less than 1.0 it is considered incomplete penetrance.
PleiotropyThe spectrum of phenotypic traits possible due to a single disease genotype.
QTL“Quantitative Trait Locus”. A locus affecting a characteristic that is produced by multiple factors and measured on a continuous scale.
Recessive AlleleAn allele that must be present in homozygous fashion to create a phenotype.
Table 2. Hearing loss genes displaying both syndromic & nonsyndromic disease
GeneLocusProtein & FunctionPhenotypesReferences
CDH2310q21-22OtocadherinDFNB12OMIM 605516
   Stereocilia connectorPresbycusis(Bork et al.2001)
   Usher 1D(Noben-Trauth et al.2003)
COL11A26p21.3Collagen, Type XIDFNA13OMIM 120290
   OSMED(Li et al.2001)
   Stickler Syndrome 
MYH922q11.2Myosin Heavy Chain 9DFNA17OMIM 160775
   HMTC(Dong et al.2005)
MYO7A11q13.5Myosin VIIaDFNA11OMIM 276903
   Actin ATPase motorDFNB2(Weil et al.1995)
SLC26A47q31PendrinDFNB4OMIM 274600
 (PDS)  Anion transporterEVA & Mondini dysplasia(Wu et al.2005)
   Pendred Syndrome 
USH1C11p15.1HarmoninUSH1COMIM 605242
   Stereocilia structureDFNB18(Ahmed et al.2002)
WFS14p16.1WolframinWolfram SyndromeOMIM 606201
   Calcium channelDFNA6/14/38(Bespalova et al.2001)
    (Young et al.2001)

In the past decade, modifier genes have been shown to produce significant alterations in both Mendelian and multifactorial processes, including hearing loss (Nadeau,2005). Due to this changing paradigm of inheritance and the significant prevalence of genetic causes of hearing loss, this article will review modifier genes and allelism in the auditory system. By focusing on two genes illustrative of this concept, we will demonstrate the complex relationship between genotype and phenotype. These concepts are shaping our current approach to diagnosing congenital hearing loss and seeking molecular therapies for hearing rehabilitation.


Modifier genes qualitatively or quantitatively alter the phenotype produced by another gene. Modifier genes have been shown to affect penetrance, dominance modification, progression, expressivity, pleiotropy, and age of onset (Nadeau,2001). Acting in a qualitative fashion, a modifier may be required for a primary gene to express a phenotype. However, it appears that modifiers act more commonly in a quantitative manner. Modifiers may often underlie quantitative trait loci (QTL) such that the summation of positive and negative modifier effects may greatly influence the resulting phenotype. This quantitative model explains how a disease allele may cause significantly different phenotypes in two families. Each individual in an outbred human population will possess a unique complement of modifiers that influence the primary disease allele to produce variable phenotypic results. If only one modifier gene was associated with the quantitative majority of a phenotype, then this would reveal a high level of correlation in linkage analyses. However, the study of modifier genes appears complicated by the large diversity of modifiers, each of which alter the phenotype in relatively minimal ways. A threshold model has been proposed in which the summation effect of all modifier genes acting on a target must cross a threshold to alter a phenotype.

A modifier gene is generally considered to be a single allele, though homozygous alleles are expected to produce a quantitatively greater effect. In addition to altering the phenotypes produced by other genes, modifier genes may independently express a phenotype. Therefore, in a complex pathway, one gene may be considered to be modified by the other genes in the pathway. Such complex interactions potentially involve multiprotein structures, intracellular processes, and intercellular communication. Therefore, the term “modifier gene” is relative to the primary gene under examination. For example, assume that we wish to study protein A. Protein A interacts in some manner with protein B. The phenotype produced by protein A differs based on its interaction with different protein B alleles, B1, B2, etc. In this example, protein B is a modifier of protein A. Note that this relational classification is further complicated by the fact that allelism in the primary protein may alter the modification of a phenotype by a modifier gene. There are multiple examples of modifiers in mice and humans (reviewed in Nadeau,2003). In mice, modifier genes are represented by the background genotype in a strain. It is well known that one disease gene may present drastically differing phenotypes ranging from a wild-type mouse without disease to incompatibility with life due to the strain-based genomic background (Nadeau,2005). This represents the significant effects of the modifier genes within the genetic background.

Modifier genes may be important in evolution. That is, modifiers may provide protection in complex functional networks in which epistasis causes the whole genome and not just a genotype to be considered as the selective unit (Nadeau,2003). It is also possible that modifiers are the result of an outbred genetic background that includes a low but significant mutational rate in a complex organism. In either case, modifiers do not appear to be highly conserved. In terms of evolutionary response, modifiers may represent the fact that “polymorphic challenges require polymorphic responses” (Nadeau,2005). Though beyond the scope of this review, modifier genes certainly alter the response to environmental exposures. For example, mutations in CDH23 will be discussed that predispose a patient not only to presbycusis, a multifactorial disease involving significant environmental insults, but also to noise-induced hearing loss (Holme et al.,2004). Noise exposure represents one of a multitude of environmental factors that show significant variability in the production of hearing loss due to genetic factors.

The past several decades have seen an explosion in gene technology. The human genome project yielded a complete sequence, allowing an unprecedented advance in the study of genomics. Likewise, complete sequencing of the mouse genome has significantly aided understanding of the genetics behind inbred mouse strains. Genomic studies begin with linkage analyses by tracking phenotypes within families. In such studies, the phenotype is linked to a genotype inherited in a classic Mendelian pattern. However, family studies lose power for multifactorial diseases and rare hereditary diseases. More recent studies have tracked phenotypes within a human population to find quantitative trait loci (QTL), or loci that contribute a quantifiable portion of a phenotype. QTL-based studies are presently only feasible with genes contributing to a large percentage of a phenotype (Haider et al.,2002). In our outbred human population, vast differences in modifier genes create a nonlinear relationship between genotype and phenotype unlike that found in inbred animal models. An outbred genetic background may contain an unidentified QTL or modifier gene that alters the phenotype under study. Different family genomes may contain multiple and different modifiers such that a single modifier cannot be tracked by phenotypic linkage analysis. Furthermore, environmental effects cause disease both independently and through interactions with hereditary susceptibility or resistance factors. Therefore, environmental effects further complicate genetic studies in an outbred population in which both environmental exposures and hereditary susceptibility to those exposures cannot be controlled.


Cadherin 23, also known as otocadherin, is part of the cadherin superfamily of cell surface adhesion proteins. Mutations in the cadherin 23 gene (CDH23) have been linked to age-related hearing loss (AHL), Usher syndrome type 1 subtype D, and a form of autosomal recessive hearing loss designated DFNB12 (Astuto et al.,2002; Noben-Trauth et al.,2003). In mice, the cadherin 23 gene, Cdh23, includes 69 exons in greater than 350 kb of genomic DNA (Di Palma et al.,2001b). The human gene CDH23 includes exons and introns very similar in size and position to the mouse, representing a high degree of conservation.

The cadherin superfamily of proteins mediate cell-cell adhesion, cell sorting, and cell migration. Each cadherin protein includes a transmembrane domain and a number of extracellular (EC) domains. The transmembrane and intracellular parts of cadherin 23 are hypothesized to interact with the actin core of a stereocilium through intermediary proteins such as harmonin b (Boeda et al.,2002). The extracellular portion of cadherin 23 includes 27 EC domains, each spanning approximately 4 nm, containing the calcium-dependent binding motifs. The EC domains function in adhesion via calcium-dependent dimerization in both parallel and antiparallel orientations (Di Palma et al.,2001b). Calcium-chelating agents disrupt cadherin binding, though the links reform approximately 1 day following chelator removal (Michel et al.,2005). Calcium-binding domains utilize 6–7 oxygen atoms in the amino acid structure and loss of a single oxygen atom can reduce the calcium-binding affinity of an EC domain to nondetectable levels (de Brouwer et al.,2003).

Cadherin 23 has been identified via in situ hybridization in cochlear and vestibular stereocilia and in Reissner's membrane (Wilson et al.,2001). Various linkages form between cochlear stereocilia, including tip, lateral, ankle, and cuticular plate links (Fig. 1) (Goodyear and Richardson,1992). Since cadherin 23 contains a relatively large extracellular domain hypothesized to have adhesive functions, it has been studied as a potential component of these stereocilia links. Furthermore, mice with homozygous mutant cadherin 23 alleles display stereocilia disarray, suggesting that cadherin 23 is functionally active in stereocilia linkages (Di Palma et al.,2001a). However, immunohistochemical studies have had contradictory results regarding whether cadherin 23 is a structural component of adult stereocilia.

Figure 1.

Linkages between inner ear stereocilia. Stereocilia have a variety of linkages including tip links, lateral links, and ankle links (Goodyear et al.,1992). The tip link has been proposed as a part of the mechanotransduction system (Sollner et al.,2004).


Cadherin expression, investigated with three separate polyclonal antibodies, showed cadherin 23 to be present along the entire length of cochlear and vestibular stereocilia at its earliest presentation during development (Boeda et al.,2002). These antibodies were specific for human cadherin 23 amino acids 1161–1174, 2456–2470, and 3324–3339. This study found that in the early postnatal days, cadherin 23 is concentrated at the distal ends of the stereocilia. By postnatal day 30, however, only a few vestibular stereocilia and no cochlear stereocilia stained for cadherin 23. This pattern of staining matched the spatiotemporal distribution of harmonin b. Harmonin b interacts with cadherin and may anchor cadherin 23 intracellularly to the actin filaments (Boeda et al.,2002). Additional immunofluorescent evidence is consistent with these results and showed cadherin 23 in developing, but not in mature, stereociliary tip structures (Lagziel et al.,2005).

In contrast to the above studies, several other studies support cadherin 23 presence in both developing and adult stereocilia. Studies utilizing a different polyclonal antibody specific for amino acids 3133–3291 showed cadherin 23 at the distal region of stereocilia in several adult species (Siemens et al.,2004; Sollner et al.,2004). Further, another recent study utilizing yet another unique antibody found similar staining patterns for cadherin 23 in stereocilia of mature mice (Rzadzinska et al.,2005). These studies also showed that loss of the tip link due to calcium chelator or elastase treatment correlated to loss of cadherin 23 staining. Loss of the tip link removes a structural support necessary for the normal stereociliary organization, producing stereociliary disarray. Mutation in Cdh23 also produces stereociliary disarray that lacks cadherin 23 staining. Finally, cadherin 23 was shown to coimmunoprecipitate with myosin 1c, a motor involved in the stereocilia mechotransduction system (Gillespie and Cyr,2004; Tsuprun et al.,2004).

Though immunohistochemistry has produced inconclusive results, functional studies, electron microscopy, and protein modeling support the hypothesis that cadherin 23 creates the tip link in mature animals. Cadherin 23 function in the mechanotransduction system may be evolutionarily conserved across species and studies in zebrafish strains with cadherin 23 mutations demonstrated stereociliary bundles with splayed distal ends suggestive of dysfunction only in the tip links, but not the lateral or ankle links (Sollner et al.,2004). The mutation that produced the mildest splaying, which affected only 5% of the stereocilia, caused a complete absence of the extracellular potential. This suggests that cadherin 23 may be involved independently in the structure of the tip links and in the mechanotransduction complex.

A related study in 2004 utilized transmission electron microscopy (TEM) to visualize the tip links and kinocilia links in chickens (Tsuprun et al.,2004). The TEM revealed a helix-like structure with 4.3 nm spacing between individual globular units. This is consistent with the hypothesized length of cadherin 23 EC domains. The authors hypothesize that two cadherin 23 molecules per stereocilia interact in a parallel manner to form a helix-like structure and that the distal amino ends of two such helices interact to produce a tip link between adjacent stereocilia. Overall, these studies suggest that cadherin 23 is part of the tip link between mature stereocilia and may be involved in other stereocilia links only during development.


Presbycusis, typically a clinical term, refers to hearing loss in the elderly human population and represents a multifactorial process (Gates and Mills,2005). Presbycusis usually presents as an increased hearing threshold in high frequencies, though it often progresses to include middle and low frequencies, causing significant hearing disability. It includes the complex interaction of hereditary influences and environmental stresses accumulated over a lifetime. Adding to the complexity, the central or peripheral auditory pathways may be affected. Schuknecht characterized presbycusis by six histopathological types, including sensory (outer hair cell loss), neural (ganglion cell loss), metabolic (strial atrophy), cochlear conductive (theoretical based on basilar membrane stiffness), mixed, and indeterminate (Schuknecht,1955; Schuknecht et al.,1993). It has been noted that 25% of cases are indeterminate and the majority of cases are considered mixed due to the inclusion of multiple histopathological elements (Scholtz et al.,2001; Gates and Mills,2005). This pathological diversity is highly suggestive of damage accumulation via multiple processes.

Cadherin 23 allelism is a quantitative factor in presbycusis. As a multifactorial disease, presbycusis includes a number of other characterized genes, many of whose products interact with cadherin 23. Due to the difficulty of studying an outbred and older human population that includes untraceable environmental insults, most studies have identified age-related hearing loss genes in laboratory mice. In mice, presbycusis is referred to as age-related hearing loss. Environmental control in the laboratory setting may exclude noise and toxin exposure from mice such that age is the most significant factor in AHL. AHL has been linked to a number of genes and is considered a multifactorial disease involving the summation of multiple QTL components. The most significant hereditary quantitative contribution to age-related hearing loss comes from the Ahl gene, which acts in a recessive manner (Johnson et al.,2000).

Recently, it was shown that the Ahl gene present in some strains is an allele of Cdh23, which encodes cadherin 23 (Fig. 2) (Noben-Trauth et al.,2003). Here we will refer to Ahl as a specific allele of Cdh23 by using the notation Cdh23Ahl. The Cdh23Ahl gene produces an in-frame splice site such that exon 7 is skipped due to a G→A transition at nucleotide 753 (Noben-Trauth et al.,2003). Exon 7 encodes 43 amino acids that form sections of the third and fourth EC domains, which is a potential homodimerization site. This mutation potentially disrupts the adhesive properties of the extracellular portion of cadherin 23 without affecting the transmembrane or cytoplasmic portions. The Cdh23Ahl gene has been characterized as a hypomorphic allele because it may both alter adhesive functions and may also involve abnormal intracellular targeting due to potential misfolding of the protein product.

Figure 2.

Mutation spectrum in cadherin 23. Cadherin 23 is a single-pass transmembrane protein that includes EC domains (27 gray bars), a transmembrane domain between the dashed lines, and a cytoplasmic domain at the carboxy terminus. Hollow markers designate one or more missense mutations. Solid markers designate deletion, insertion, frameshift, and nonsense mutations. In the human and the mouse genes, the majority of identified mutations are in the EC domains. Human mutations are specific to families and divided by disease phenotype (Astuto et al.,2002). In the mouse, mutations are denoted by waltzer (v) subtype or by the Ahl gene denoting Cdh23Ahl found in many inbred strains (Di Palma et al.,2001a; Noben-Trauth et al.,2003). All waltzer subtypes are believed to be functionally null mutations (Libby et al.,2003).

Mouse strains provide an exceptional means to study Cdh23Ahl and its modifiers by maintaining a strain-specific genomic background. The mouse strain C57BL/6J and more than nine other strains carry the Cdh23Ahl gene in a homozygous manner (Erway et al.,1993; Johnson et al.,2000). These strains experience hearing loss occurring first in the high-frequency range that progresses over time to become profound deafness across all frequencies. Both the time of hearing loss onset and the rate of progression for Cdh23Ahl homozygotes are related to the strain-based genomic background (Johnson et al.,2000). For instance, the C57BL/6J strain of mice has measurable high-frequency hearing loss beginning around 2 months of age, though it does not become significant until after 10 months (Li and Borg,1993). Another strain of mice carrying the Cdh23Ahl allele in a homozygous manner is DBA/2J (Erway et al.,1993). The DBA/2J strain experiences a faster progression of AHL in comparison to the C57BL/6J strain presumably due to strain-specific gene differences.

Cdh23Ahl-related functional hearing loss is associated with a number of pathological changes that begin in the basal cochlear turns (Henry and Chole,1980). The histopathological progression of hearing loss begins with stereocilia disarray in young mice that subsequently becomes cochlear hair cell loss and spiral ganglion degeneration in older mice (Di Palma et al.,2001a). Interestingly, a digenic mouse displaying heterozygosity for both Cdh23Ahl and a mutant Pcdh15 shows stereocilia disarray and hearing loss first measurable at 5 months (Zheng et al.,2005). Protocadherin 15 is part of the cadherin superfamily and involved in stereocilia structure, likely forming external linkages similar to cadherin 23. Mutations in the protocadherin 15 gene, Pcdh15, may cause Usher syndrome type 1. Protocadherin 15 also appears to be a modifier of Cdh23. Protocadherin 15 is hypothesized to be one of several proteins that interact with cadherin 23 in the formation of a macromolecular complex required in stereocilia development.

A number of genes or loci have been identified as Cdh23 modifiers, including several mutant mitochondrial alleles, ahl2, ahl3, and Atp2b2 (reviewed in Noben-Trauth et al.,2003). These particular modifiers have been shown to add to the severity of the hearing loss when combined with homozygous Cdh23Ahl alleles. These modifier genes have been shown to cause independently a hearing loss phenotype, though they were identified later than Cdh23Ahl because of their smaller quantitative contributions to a hearing loss phenotype.

The modifier of deaf waddler (mdfw) is allelic to Cdh23Ahl and both are found to be severity modifiers of hearing loss caused by mutation of the Atp2b2 gene, which encodes a calcium pump designated PMCA2 (Zheng and Johnson,2001). Reciprocally, PMCA2 has been shown to be a modifier of CDH23 in humans by causing more severe hearing loss (Schultz et al.,2005). PMCA2 has been localized to the distal regions of hair cell stereocilia and to the spiral ganglion (Yamoah et al.,1998; Kozel et al.,2002). Its distal location and modifier associations with CDH23 suggest that PMCA2 may produce an enriched calcium microenvironment required for cadherin 23-containing stereocilia links.

Though modifiers of Cdh23 may independently produce hearing loss, the relationship between the collective genomic effects and the phenotype is likely to be nonlinear. Genetic components of hearing loss may experience a threshold effect such that multiple additive and subtractive modifiers are summed to represent the genetic background effects. In the case of AHL and presbycusis, only a sum above a threshold level will allow environmental insults (e.g., age, noise, toxic agents) to produce the hearing loss phenotype. Further, hereditary sensitivity or resistance may be specific for a single type of environmental insult.


Nonsyndromic recessively inherited sensorineural hearing loss associated with mutations in cadherin 23 is termed “DFNB12.” DFNB12 was mapped in a Sunni family to chromosome 10 and later found to be allelic with Usher 1D (Chaib et al.,1996; Bork et al.,2001). Although mutations in CDH23 may cause either DFNB12 or Usher syndrome type 1, the nonsyndromic hearing loss is generally of less severity (Pennings et al.,2004). DFNB12 produces a moderate to profound hearing loss that is diagnosed by 6 years of age, usually nonprogressive, and generally more severe in the higher frequencies (Astuto et al.,2002; Pennings et al.,2004). DFNB12 is characterized as a nonsyndromic disease, but asymptomatic retinitis pigmentosa-like lesions in some cases suggest a relative continuum of phenotypes between DFNB12 and Usher 1D due to Cdh23 allelism (Astuto et al.,2002).

Studies have shown that all characterized DFNB12 cases are due to missense mutations in the highly conserved calcium-binding domains of cadherin 23 (Fig. 2) (Astuto, et al.,2002). An interesting mutation associated with DFNB12 destroys the secondary structure by affecting a salt bridge and therefore may affect the structure of the EC domains in a similar fashion to an inherent EC mutation (de Brouwer et al.,2003). In DFNB12 and other nonsyndromic recessive hearing loss disorders, the genotype may be homozygous, compound heterozygous, or digenic with a single heterozygous allele and a second mutation in another gene (Astuto et al.,2002). Digenic heterozygous mutations in CDH23 (DFNB12) and GJB2 (DFNB1) have been shown to produce hearing loss clinically similar to DFNB12 (de Brouwer et al.,2003). Overall, DFNB12 involves homozygous or compound heterozygous missense mutations affecting the calcium-binding properties of the EC domains. There are presently no mouse or other animal models of DFNB12. Present literature does not contain any histopathological characterization of the inner ear in DFNB12-affected individuals. We assume that DFNB12, as a more severe form of hearing loss than presbycusis, will also display stereocilia disarray.

In summary, presbycusis and DFNB12 are associated with missense mutations in the EC domains. These missense mutations all occur in highly conserved sites involved in calcium-dependent adhesion such that the cadherin 23 extracellular domain is present but less functional. Rzadzinska et al. (2005) recently showed that levels of cadherin 23 expression are equivalent in mouse strains that are sensitive or resistant to age-related hearing loss. This supports the hypothesis that normal expression levels, but of a dysfunctional cadherin 23, causes loss of stereocilia linkages, most likely tip links, resulting in nonsyndromic hearing loss. The transmembrane and intracellular moieties of cadherin 23 may be unaffected and able to function within the putative macromolecular complex, allowing normal stereocilia morphogenesis.


“Usher syndrome” is an umbrella term for a family of autosomal recessive diseases that include both sensorineural hearing loss and visual impairment related to retinitis pigmentosa. Charles Usher (1916), a Scottish ophthalmologist, observed retinal pigment disorders concurrently with hearing loss occurring in a familial pattern. Usher syndrome is divided into three types by clinical severity and components. Usher syndrome type 1 includes congenital, profound, nonprogressive deafness, vestibular areflexia, and retinitis pigmentosa. Type 2 includes moderate to severe sensorineural hearing loss and retinitis pigmentosa. Type 3 includes progressive hearing loss, variable vestibular disease, and retinitis pigmentosa (Keats and Savas,2004). Usher syndrome in its various types affects an estimated 50% of concurrently deaf and blind individuals (Boughman et al.,1983), 18% of retinitis pigmentosa-affected individuals, and 3–6% of congenitally deaf individuals (Vernon,1969).

Usher syndrome type 1 has been associated with a number of genes, with MYO7A and CDH23 being the most common subtypes (Keats and Savas2004; Ouyang et al.,2005). The five identified genes causing subtypes of Usher type 1 are MYO7A (Usher 1B), USH1C (Usher 1C), CDH23 (Usher 1D), PCDH15 (Usher 1F), and SANS (Usher 1G) (Boeda et al.,2002; Keats and Savas,2004). Aside from causing Usher syndrome, each of these genes may be a modifier of CDH23 in hearing loss due to functioning in stereocilia structure or morphogenesis (Fig. 3). Histopathologic findings in Usher syndrome type 1 are consistent across four patients (Wagenaar et al.,2000). Usher 1 histopathologic features include degeneration of the organ of Corti and cochlear nerve, atrophy of the stria vascularis, decrease in spiral ganglion cell number, and sometimes degeneration of the inferior vestibular nerve.

Figure 3.

Interactions of proteins associated with Usher syndrome type 1. This image shows conceptual protein interactions in inner ear and retinal stereocilia. Black arrows denote interactions occurring within the stereocilia. Circular arrows denote homodimer formation. Gray arrows denote sans interactions, though these likely occur external to the stereocilia (Adato et al.,2005).

Knowledge of Usher syndrome type 1 in humans has advanced significantly due to mouse models containing mutations in the subtype gene orthologs. Usher syndrome type 1 by definition includes vestibular dysfunction. Usher 1-associated mouse models display vestibular dysfunction in obvious abnormal movements that have earned phenotype-based names such as waltzer, circler, jerker, and shaker (Table 3). The waltzer phenotype is homozygous null in Cdh23 via nonsense, frameshift, or splice-site mutations (Di Palma et al.,2001b). Waltzer stereocilia and kinocilia show disorganization in both the cochlear and vestibular systems (Di Palma et al.,2001a; Lagziel et al.,2005).

Table 3. Mouse models of Usher syndrome type 1
SubtypeMouse StrainMutant AlleleProteinReference
  • *

    Harp has not been definitively identified as the product of SANS (Johnston et al.2004).

Usher 1BShaker-1Myo7ash1Myosin VIIa(Gibson et al.1995)
Usher 1CDeaf CirclerUsh1cdfcrHarmonin(Johnson et al.2003)
Usher 1DWaltzerCdh23vCadherin 23(Bolz et al.2001)
Usher 1FAmes WaltzerPcdh15avProtocadherin 15(Alagramam et al.2001)
Usher 1GJackson ShakerSansjsSans(Weil et al.2003)
   Harp*(Johnston et al.2004)

Usher syndrome 1D in humans is orthologous to the waltzer mouse models and includes more severe types of mutations in Cdh23 (Zheng et al.,2005). In one series, Usher 1D mutations were found to be null alleles in 88% of cases (Astuto et al.,2002). All waltzer alleles of Cdh23 appear to be functionally null (Libby et al.,2003). A small percentage of CDH23 mutations causing Usher 1D has been identified as missense mutations in the EC and cytoplasmic domains (Astuto et al.,2002; Ouyang et al.,2005). It is expected that these missense mutations either produce a functionally null cadherin 23, or that unidentified severe mutations reside in other Usher syndrome type 1 genes in these families.

Recessive null mutations in CDH23, MYO7A, PCDH15, SANS, or USH1C are known to produce the Usher syndrome type 1, which is the most severe form of Usher syndrome producing profound deaf-blindness and vestibular dysfunction (Table 4). This is in stark contrast to the functional missense mutations in CDH23 associated with the less severe DFNB12 phenotype. This suggests that Usher type 1 is caused by dysfunction of a complex of stereocilia proteins required in both development and function (Frolenkov et al.,2004). The complete loss of any component protein by a null mutation may inhibit formation of the entire complex. This severe mutational effect would then block stereocilia morphogenesis in all related tissues, resulting in syndromic disease. In the case of Usher syndrome type 1, the macromolecular complex appears to be necessary for retinal development and stereocilia development in the inner ear.

Table 4. Web resources in human and mouse genomics
Hereditary hearing impairment in mice
Hereditary hearing loss homepage (human)
Human genome map
MITOMAP (human mitochondrial genome)
Mouse genome map
OMIM - Online Mendelian Inheritance in Man
WFS1 Mutation and polymorphism database

Multiple studies have attempted to characterize both the spatiotemporal aspects of cadherin 23 and those proteins with which it interacts. Present knowledge suggests that cadherin 23 interacts directly or indirectly with harmonin b, myosin 7a, myosin 1c, and SANS (Fig. 3) (Boeda et al.,2002; Weil et al.,2003; Frolenkov et al.,2004; Siemens et al.,2004; Adato et al.,2005). These proteins have been shown to interact in stereocilia morphogenesis and function. Cadherin appears to bind to the actin core via a harmonin b anchor. Protocadherin 15, myosin 7a, and SANS also interact directly with harmonin b. The present model involves transportation of harmonin b to proper stereocilia regions by the actin motor, myosin 7a. Cadherin 23 appears to match harmonin b location at various developmental stages. The SANS protein has not been found inside stereocilia, but has been localized to the proximal end of the cell in the pericuticular plate. At this location, SANS has been shown to interact with harmonin, myosin 7a, and protocadherin 15. Based on this location, SANS may regulate or facilitate the entrance of the other proteins into the stereocilia and thereby be a necessary component of the macromolecular complex formation (Adato et al.,2005).

Further studies of the interactions within the putative macromolecular complex may lead to genetic identification of the remaining subtypes of Usher syndrome type 1. For example, it was recently shown by using a yeast two-hybrid system that a novel integral membrane protein, PHR1, interacts with myosin 1c and myosin VIIa (Etournay et al.,2005). Creation of a mouse PHR1 knockout may produce an Usher type 1 phenotype if the PHR1 product is only involved in a protein framework for stereocilia morphogenesis and function that involves the other Usher 1 proteins. Myosin 1c might also be an Usher 1 candidate protein due to its interaction with cadherin 23. However, myosin 1c is involved in neural cone turning (Wang et al.,2003), cell motility, macrophage phagocytosis, and stereocilia structure in the inner ear and retina (Barylko et al.,2005). Null mutations in myosin 1c are likely incompatible with life, although an inner ear-specific knockout mouse may display the same phenotype as the waltzer strain.


As shown in the above examples, CDH23 allelism produces three recognized phenotypes. These disorders represent a spectrum of disease severity that roughly correlates to progressively greater dysfunction in cadherin 23 due to allelism and modifier effects. Complete loss of cadherin 23 due to a null mutation has been identified as the cause of Usher 1D in the majority of cases. In contrast, missense mutations particularly within the extracellular domains of cadherin 23 produce a complete, or near complete, protein with reduced function that is associated with DFNB12 or presbycusis. Supporting the hypothesis of a continuum of phenotypes are the examples that fall in between the main phenotypes. For instance, DFNB12 could be considered a syndromic disease due to its subclinical RP-like manifestations that may cause night blindness at their most severe presentation (Astuto et al.,2002). However, CDH23 allelism yields part of the explanation for the variable phenotypic presentation. Beyond these impressively diverse allelic effects, modifier genes appear to affect the phenotypic expression and likely represent specific proteins interacting with cadherin 23 in stereocilia function. Given the complex set of spatiotemporal interactions in stereocilia development and function, it is likely that the putative macromolecular complex interacts significantly with modifier gene products.


Hereditary hearing loss due to mutations in the WFS1 gene takes two known forms: autosomal dominant nonsyndromic and autosomal recessive syndromic hearing impairment. The autosomal dominant, nonsyndromic hearing loss, termed “DFNA6/14/38,” produces congenital hearing loss specifically in the low-frequency range. This is one of only two conditions identified exhibiting nonsyndromic low-frequency hearing loss. DFNA6/14/38 has been described in separate families as either progressive or nonprogressive (Pennings et al.,2003). Wolfram syndrome is due to recessive mutations in the WFS1 gene. This syndrome is also known by the acronym DIDMOAD for its cardinal phenotypic components, which include diabetes insipidus, diabetes mellitus, optic atrophy, and deafness (Strom et al.,1998; Philbrook et al.,2005). The deafness component of Wolfram syndrome occurs in 62% of patients and predominantly affects the high-frequency range (Pennings et al.,2004). In most cases, patients will present with a nonautoimmune diabetes mellitus during their first decade and optic atrophy during their second. However, this syndrome is highly variable in presentation and may include neural degenerative features, pituitary-gonadal axis dysfunction, or a predisposition to psychiatric illnesses (Philbrook et al.,2005). In general, Wolfram syndrome produces a progressive neurodegenerative disorder that affects central, peripheral, and neuroendocrine components of the nervous system and results in death by age 40.


The WFS1 gene is on chromosome 4p16 and spans 8 exons. This gene encodes an 890 amino acid protein called wolframin that is hypothesized to contain nine transmembrane regions and has been shown to have an Ncyt/Clum orientation (Hofmann et al.,2003). Protein monomers are approximately 100 kDa but wolframin elutes in an approximately 400 kDa fraction from cellular extracts, suggesting the functional formation of a homo-oligomer. Based on mRNA hybridization studies, wolframin is expressed in a number of tissues, including heart, brain, placenta, lung, and pancreas (Strom et al.,1998). It has also been demonstrated in insulinoma β-cell lines (Hofmann et al.,2003). Intracellularly, the majority of wolframin colocalizes with calreticulin, suggesting a major functional role in the endoplasmic reticulum (ER) (Cryns et al.,2003b). Expression of wolframin in murine inner ear development was shown by immunohistochemistry and in situ hybridization to involve the canalicular reticulum (Cryns et al.,2003b). Canalicular reticulum is a form of ER similar to the tubulocisternal ER located in some ion-transporting epithelia (Cryns et al.,2003b). The canalicular reticulum has been identified in the cells of various species other than mice, including vestibular and cochlear hair cells, spiral ganglion, Deiter's cells, and the stria vascularis. Contrary to functional hearing losses showing high-frequency loss in Wolfram syndrome and low-frequency loss in DFNA6/14, immunohistochemistry and expression studies have not been able to identify any differences between the apical and basal cochlea in relation to wolframin expression (Cryns et al.,2003b).

Wolframin localization to canalicular reticulum-containing cells suggests the transport of either potassium or calcium, both of which are important in these cells (Hofmann et al.,2003). Indeed, wolframin has been shown to control calcium homeostasis, putatively as either a channel or regulator (Osman et al.,2003). Calcium transport via ER is intimately involved in apoptosis (Strehler et al.,2004). Murine pancreatic cells containing a mutant Wfs1 showed an increase in sensitivity to ER stress via tunicamycin and thapsigargin (Ishihara et al.,2004). These cell isolates underwent increased levels of apoptosis in response to either high glucose or ER stress inducers, though these results were not repeatable 1 year later (Philbrook et al.,2005).

Alternative hypotheses suggest that wolframin is involved in regulation of the cell cycle and proliferation. A recent study analyzed the expression levels of 10,000 cDNA species between one Wolfram patient and a control patient utilizing DNA hybridization array technology (Philbrook et al.,2005). The Wolfram patient cells showed upregulation of fibulin-3, which is associated with cellular senescence. In agreement with fibulin-3 function, antisense-based repression of wolframin expression inhibited proliferation but did not increase the rate of cell death (McBain and Morgan,2003). Further, comparing primary fibroblasts from Wolfram patients and normals showed no difference in response to irradiation, suggesting that wolframin does not modulate these processes through DNA repair mechanisms (Philbrook et al.,2005). These studies suggest that wolframin is involved in calcium transport, though it is unclear how this relates to its apparent effects on cellular senescence or apoptosis.


Similar to CDH23, mutations affect the WFS1 gene in an allelic manner between Wolfram syndrome and DFNA6/14/38. DFNA6 and DFNA14 loci were initially identified as close but distinct. However, DFNA6 was later shown to be allelic with DFNA14 in the WFS1 gene, leading to the designation “DFNA6/14” (Bespalova et al.,2001). In that same year, the DFNA38 locus was inconclusively associated with a WFS1 mutant allele, resulting in the designation “DFNA6/14/38” (Young et al.,2001). Isolated mutations in WFS1 that produce hearing loss are noninactivating mutations that aggregate in the exon 8 C-terminal protein domain (Cryns et al.,2003a; Gurtler et al.,2005). These heterozygous WFS1 mutant alleles include missense mutations and one in-frame three base pair deletion (Cryns et al.,2003a). Following age-based correction for presbycusis using ISO 7029, only a portion of families with a heterozygous mutation in WFS1 were shown to experience hearing loss (Pennings et al.,2003). This suggests that a DFNB6/14/38 patient may present congenitally with a nonprogressive hearing impairment that leads to better hearing in comparison to presbycusis-affected peers in later life.


There is a spectrum of mutations within the WFS1 gene associated with Wolfram syndrome. Wolfram syndrome-associated mutations span the entire gene (Cryns et al.,2003a). The majority of mutations are nonsense, frameshift, or significant insertions or deletions expected to produce a functionally null protein. However, a significant proportion of the mutations in Wolfram syndrome are missense or small in-frame insertions or deletions expected to produce a functional protein. In one series, nonsense mutations in Wolfram syndrome individuals produced mRNA transcripts that were not expressed due to nonsense-mediated messenger decay (Hofmann et al.,2003). In the same report, a missense mutation in the fifth extracellular domain caused protein degradation due to instability. This led to a significantly reduced wolframin half-life, in which only 10% of the expected protein concentration was produced by the missense allele. The Wolfram syndrome patient with the missense mutation in WFS1 was a compound heterozygote in whom the second allele was a nonsense mutation that experienced nonsense-mediated messenger decay. Thus, a transcript containing a nonsense mutation will undergo degradation prior to expressing any protein, whereas a missense mutation will lead to significantly reduced but measurable protein levels due to wolframin instability. This suggests that the spectrum of phenotypes produced by WFS1 mutant alleles is largely due to reduced protein dosage.

Further support for protein levels as a determinant of phenotype comes from a report of two female patients from the same family who were compound heterozygous in WFS1 with noninactivating missense mutations (Pennings et al.,2004). These patients presented with the mild syndromic feature of diabetes mellitus and no hearing loss. In this case, missense mutations that were noninactivating, and also not degraded as nonsense, caused a mild and limited form of the syndromic disease. In fact, homozygous or compound heterozygous missense mutations lead to a mild phenotype in 50% of cases (Cryns et al.,2003a). In contrast, compound heterozygotes in WFS1 in which one allele had a missense mutation, and the remaining allele was inactivating (i.e., reduced functional protein levels), led to a mild phenotype in only 7% of cases.

Mutant WFS1 alleles have been described in Spanish, British, Dutch, Italian, and Turkish Wolfram syndrome families (reviewed in Domenech et al.,2004). In Spanish families, the most common mutant allele involved a duplication in exon 4, though all other mutations were identified in exon 8. The majority of the patient alleles were homozygous. In contrast, a study of nine American families also identified most mutations in exon 8, although in all but one consanguinous family the WFS1 genotype included compound heterozygous alleles (Smith et al.,2004). To date, there have been no conclusive genotype-to-phenotype associations in Wolfram syndrome. Syndromic mutations are homozygous or compound heterozygous and involve inactivating mutations of the WFS1 gene in the majority of cases.


Allelism in the WFS1 gene produces a spectrum of disease severity and pleiotropy ranging from nonprogressive nonsyndromic hearing loss to severe forms of Wolfram syndrome that may include progressive and profound hearing loss. Though the function of both wild-type and mutated wolframin has yet to be characterized, there is a significant difference between the type and location of the mutations that lead to nonsyndromic and syndromic hearing loss. DFNA6/14/38 mutations cluster in the C-terminal region, whereas Wolfram syndrome mutations are spread out along the entire gene. It is hypothesized that Wolfram syndrome occurs due to decreased wolframin levels, but DFNA6/14/38 occurs due to a restricted loss of function in wolframin. That is, inactivating mutations of WFS1 do not produce any wolframin protein and thus cause a Wolfram syndrome phenotype. Noninactivating mutations in regions of WFS1 earlier than exon 8 may not be well tolerated in terms of protein folding or intracellular localization. This could greatly reduce the stability of such proteins, thus reducing protein dosage also leading to a Wolfram syndrome phenotype (Hofmann et al.,2003). However, missense mutations characteristic of DFNA6/14/38 occur almost exclusively within the C-terminal domain. It may be that missense mutations are better tolerated in this region such that protein stability is not greatly reduced and normal levels of a dysfunctional protein are generated.

Of significant interest is why one type of mutation produces a dominant hearing loss in the low frequencies, whereas another produces recessive and syndromic hearing loss in the high frequencies. Microscopic studies have found no difference in the cochlear position of WFS1 expression (Cryns et al.,2003b). Though no WFS1 modifier genes have been identified, the extremely broad range of expressivity, pleiotropy, and disease progression suggests a complex set of tissue-dependent interactions. There is a small but significant percentage of Wolfram syndrome patients in whom a WFS1 mutation was not found, suggesting the presence of an unidentified mutation in a WFS1 regulatory or intronic sequence site or in a modifier gene. Supporting the hypothesis that genetic and environmental modifiers affect WFS1, sex-related differences affect the hearing phenotype in Wolfram syndrome patients (Pennings et al.,2004). In a small sample of seven families involving inactivating WFS1 mutations, four females were shown to have significantly worse hearing impairment than seven males with some sibling comparisons included. The authors suggest that sex hormones act as a modifier of the hearing loss in Wolfram syndrome, though a specific mechanism has not been identified. Further work is required to understand the function of the wolframin protein. In similar fashion to CDH23, advances in WFS1 knowledge may follow from characterization of tissue-specific interactions involving wolframin. Mouse models will likely be essential in isolating phenotypic effects and linking them to WFS1 mutations and modifier alleles.


We have reviewed recent progress in understanding genotype-phenotype relationships of two genes. Modifier genes add a level of complexity to these allelic interactions. In both CDH23 and WFS1, the phenotypic spectrum appears to be more of a quantitative continuum than several discrete entities. Given the expected ubiquitous nature of modifier genes in complex organisms, they are likely to affect all Mendelian or multifactorial hereditary phenotypes to some extent. Modifiers will blur the sharp lines between discrete and allelic phenotypes.

The rate of gene discovery in all loci and in hearing loss loci is not significantly advancing (Gao et al.,2004; Flint et al.,2005). In 2001, it was noted that 51 genes had been characterized in hearing loss over the prior 16 years (Resendes et al.,2001). Only a slight majority of those 51 genes were found in the latter half of that time period. The majority of genes characterized were primary disease genes displaying Mendelian inheritance (Table 4). Such genes showed significant linkage patterns within human families. However, modifier genes and QTL that affect only a small proportion of a phenotype will be missed in the genetic noise of most linkage studies.

Though modifier gene discovery is a slow process, important progress has been made in defining their role in hereditary hearing loss. In tubby mice, the autosomal recessive genotype (tub/tub) causes a variety of phenotypic changes, including adult-onset insulin resistance-associated obesity and early onset cochlear and retinal degeneration (Ikeda et al.,1999). A modifier gene named moth1 for modifier of tubby hearing has been found to protect hearing in a dominant fashion or worsen hearing in a recessive fashion, depending on the moth1 allele. As previously mentioned, the deaf waddler mouse contains a recessive mutation in dfw, which encodes the calcium pump Atp2b2 (Kozel et al.,2002). Recently, a modifier of dfw termed “mdfw” was not only identified but shown to be allelic for the Ahl gene on chromosome 10 (Bryda et al.,2001; Zheng and Johnson2001). Similar to moth1, the modifier mdfw displays allelism by being protective of hearing in a dominant fashion or permitting hearing loss in a recessive fashion, depending on the specific mutation.

Fewer modifiers of hearing have been identified in humans than in mice. A form of autosomal recessive nonsyndromic hearing loss termed “DFNB26” was first identified in a large consanguineous Pakistani family (Riazuddin et al.,2000). A modifier gene was found in seven family members containing homozygous DFNB26 alleles who displayed normal hearing. The modifier, termed “DFNM1,” has yet to be fully characterized (Riazuddin et al.,2000). This created a new terminology for genetic modifiers of hearing loss, DFNM, akin to designations used in recessive (DFNB) and dominant (DFNA) hereditary hearing loss. The majority of identified modifier genes in hearing loss affect the mitochondria. For example, a transcription factor and two enzymatic modifiers of mitochondrial RNA were shown to affect the well-characterized A1555G mtDNA mutation associated with aminoglycoside-induced ototoxicity (Bykhovskaya et al.,2004a,2004b). As a nonmitochondrial example, a single mutation in MYO7A was shown to produce DFNA11 of varying severity between families strongly suggesting the effect of a modifier gene (Street et al.,2004).

Characterizing modifier genes will lead to a more intricate knowledge of biological pathways and may result in better treatment options for hearing loss patients and their families (Haider et al.,2002). The characterization of modifiers may define treatment subgroups or yield completely new diagnostic and therapeutic targets. Recognition of the importance of modifiers may force new advances in the gene discovery frontier and reaccelerate the rate of identification of hearing loss-associated genes.