4. Discussion

In this paper, we have sought to identify patterns in OI‐related HL (type, degree, age of onset, and progression) as well as association of those features with OI skeletal phenotype (Sillence type) and causative gene variants. To do this, we have utilized the extraordinary longitudinal data from protocol participants with OI Type III/IV who were followed from childhood. All participants have a heterozygous causative variant in COL1A1 or COL1A2, most of which result in a glycine substitution and a small number with skipping of a single exon or a small deletion. The most significant of the novel findings resulting from this analysis is an association of progression to more severe HL with COL1A1 variants, particularly in participants who also have OI Type III.

The significant advantages of multidecade longitudinal data also allowed our analysis to amplify recent cross‐sectional reports, most notably the prevalence of early onset OI‐related HL in the first decade of life. This contrasts with earlier reports that established the orthodoxy that OI‐related HL had its onset in the second and third decades. However, our pediatric emphasis limits determination of adult‐onset HL in more recently enrolled patients, which warrants a complete characterization of HL findings over the lifespan.

In our longitudinal cohort, more than half of participants had at least mild HL, and the prevalence of HL in OI Type III was significantly greater (1.6x) than in OI Type IV. Almost all HL began as CHL and remained mild and stable in most OI Type III and OI Type IV participants over time, with about one‐quarter of those with CHL transitioning to MHL. When HL progressed to moderate or greater, the majority (67%), but not all, transitioned to MHL. This is a greater extent of HL than reported from the North American OI BBRC, with HL in about one‐quarter each of OI Types I, III, and IV [9]. The European collaboration had a strong focus on OI Type I (83.7% of participants) with an overall mean age of evaluation of 30 years. They reported an overall HL prevalence of 48.4%, with almost all converting to MHL or SNHL by the fifth decade [18]. An earlier study of a Finnish OI adult population, again with OI Type I most common, found HL in 57.9% of the patients, beginning in the second to fourth decades and most often MHL [6]. Thus, our study detected a higher percentage of HL among OI Type III/IV than the BBRC study, more in line with European studies. Given that the European studies have a majority composition of Type I OI, our data support a similar overall incidence of HL resulting from both changes in collagen primary sequence and insufficiency of normal collagen. In our study, the incidence of conversion of CHL to SNHL is low and MHL only occurs in the subgroup with progression to more severe HL. The low SNHL conversion in our group may be related to age; our cohort was predominantly under the age of 40 years, whereas other study populations skew older (40% of BBRC population over age 40 years) [6, 9]. Kuurila et al. [6] reported CHL limited to patients in the first three decades, with prevalence of SNHL increasing with age. In the BBRC report, the transition to MHL or SNHL occurs predominantly in OI Types I and III after the three decade and affects about half of these patients overall [9].

It had become “dogma” that OI‐related HL had its onset late in the second and mostly in the third decade of life. In line with these reports, our protocol initially undertook repeated audiological evaluation of children with the goal of finding features predictive of later HL development. Instead, we demonstrated onset of HL in the first decade of life in the majority of our OI Type III/IV population and before school‐age in 46% of OI Type III and 33% of OI Type IV, including among the group with progressive HL. There had been hints of early HL in other studies, likely limited by availability of children and predominance of OI Type I in study groups. In the Finnish pediatric studies, only 2 of 45 children were found with HL, both with OI Type IV, at the ages of 11 and 15 years [12]. From the European study [18] of 184 OI patients, 83.7% with OI Type I, a median age of onset of 20 years (range 5–60 years) was reported. At the younger end of this group, HL was detected in the first and second decades of life in 6.3% (n = 2/32 ears) and 34.1% (n = 30/88 ears), respectively [18]. The more recent cross‐sectional BBRC study detected HL in the first decade among approximately one‐fifth of OI Type III ears but not in OI Type IV, where onset occurred among 10% in the second decade [9]. The availability of longitudinal data likely contributed to the greater sensitivity of detection of early loss in our study, which was also drawn from across North America.

In addition, it is interesting to compare secondary features of OI Types III/IV, with heterozygous alterations in collagen protein sequence, with OI Type I, with insufficiency of normal type I collagen. There is a distinction in pathogenic mechanism and bone matrix composition between OI caused by haploinsufficiency variants and missense variants. In OI, haploinsufficiency refers to heterozygous variants in (almost always) COL1A1 that result in a lack of stable alpha chain protein synthesis by the mutant allele (generally stop codons or PTCs from out‐of‐frame splicing). The patient thus synthesizes half the normal total amount of the alpha1 chain, and all the alpha1 chain that is synthesized has normal primary structure. At the matrix level, the amount of collagen is reduced and the pathology is due to insufficiency of the heterotrimeric Type I collagen molecule (_α_1(I)2, _α_2(I)1). Individuals whose OI results from this type of mutation have the mildest form of OI, Type I OI. Many of the previous audiology studies were dominated by adults with Type I OI, representing only the clinical consequences of that pathology.

Mild HL in the first decade was reported among 19% of individuals with OI Type I without molecular confirmation [10]. The Finnish, European and BBRC studies all found that OI Type I populations had HL onset in the second to fourth decades, with the percentage of affected individuals increasing over time, supporting later development of HL in haploinsufficiency [6, 9, 18].

The availability of longitudinal data in a fully genotyped study population also yielded insights into genetic features among those with heterozygous variants in their Type I collagen primary protein sequence. Heterozygous single amino acid substitutions (almost always Gly substitutions in the first position of the Gly–X–Y collagen trimer) have a dominant negative pathologic mechanism, rather than a loss‐of‐function. The alpha chains containing the altered primary sequence are synthesized, incorporated into collagen heterotrimer in the cell, secreted and incorporated into extracellular matrix. For missense variants in one COL1A1 allele, 75% of Type I collagen is expected to contain at least one chain with abnormal primary structure, whereas for missense variants in one COL1A2 allele, 50% of Type I collagen is expected to contain a variant alpha chain. In general, missense variants are more deleterious to bone function than a reduced quantity of normal matrix. Hence, these missense variants with dominant negative pathology cause Types III OI, the severe progressive deforming type, and IV OI, which is moderately severe. The missense variants act by altering normal intracellular, cell‐matrix, matrix, and bone tissue functions. All patients in our study fall into this dominant negative category of OI pathology.

Genotype–phenotype correlations for skeletal aspects of OI have shown a complex mix of features—the proportion of Gly substitutions causing OI Type II, the perinatal lethal form, is greater for COL1A1 than COL1A2 variants, but there is not a straightforward association [30]. One possible explanation is the stoichiometry of Type I collagen heterotrimers, with two _α_1(I) and one _α_2(I) chains. Also, alpha chain functions differ, with _α_1(I) being especially critical to trimer stability and interaction with matrix molecules through a major ligand binding region, whereas _α_2(I) interactions along the chain involve matrix proteoglycans. These explanations support productive research but are not sufficient for clinically relevant predictions for patients. Here, we find that HL is associated with heterozygous variants altering the collagen primary proteins sequence (Gly substitutions, splicing, deletions) in either alpha chains, and that these variants causing HL are located along the helical region and even into the C‐propeptide (Figure 3). There is an intriguing midhelical gap in HL, which awaits reports of additional variants associated with HL from other studies to advance or undermine its functional status.

Our genetic data yielded two additional features. First, when looking at the interfamilial and intrafamilial concordance of HL outcome, we find the same phenotypic variability for HL as is found for skeletal features, with some variants having consistent outcomes and others having a variable outcome, pointing to modifying genetic factors yet to be identified. This variability was previously noted in the European study [18]. Second, and importantly, we find that, among individuals with heterozygous collagen missense variants as the cause of their OI, progression of HL beyond stable mild loss is limited to COL1A1 mutations. The progressive HL begins as CHL and generally converts to MHL. Moreover, statistical analysis of interactions between HL, gene, and OI type reveals that progression is significantly associated with COL1A1 variants causing OI Type III. This statistical preponderance of HL progression supports its association with greater overall OI severity, although noncollagen modifying factors are still likely to be contributory. An association of lower areal lumbar vertebral DXA z‐scores in OI patients with HL versus OI patients without HL was reported by Swinnen et al. in a genetically verified European OI group with 80% haploinsufficiency (32.7% normal hearing) and 20% collagen heterozygous missense variants (43% normal hearing) [8]. Since all haploinsufficiency mutations lead to the same outcome at the collagen protein level, undetermined modifying factors are again implied. The Finnish study found no association of HL with collagen alpha chain, but the predominantly Type I OI population and lack of longitudinal data did not allow examination of progression for individuals [7]. The BBRC study found more HL in Type III than IV OI but did not report collagen variants [9].

The data presented here extended the prevalence and severity of OI‐related HL to the molecular and genetic level but does not clarify the mechanism of its variability. At a molecular (biochemical) level, our data on HL progression in patients with COL1A1 variants are consistent with collagen stoichiometry. At the bone tissue level, CHL and MHL could result from bone fragility and bone remodeling around the otic capsule due to abnormal collagen structure and bony changes in the ossicles and otic capsule, but these anomalies are also evidenced in individuals with no HL [31]. The substantial incidence of HL in OI Type I, in which there is insufficiency of normally structured collagen, shows that abnormal collagen structure is not necessary to HL. CHL and MHL in the OI population have often been attributed to otosclerosis and stapes footplate fixation [32] but in OI, there is a higher prevalence of footplate thickening and brittleness [33].

Early identification of HL allows individuals to seek treatment through amplification and accommodations at home, school, and work [34]. Late‐identified and untreated HL has multiple impacts on language development, education, employment, cognition, mental health, and overall quality of life across the lifespan [34]. Even children with mild HL are at risk for delays in speech and language development, listening comprehension, and academic achievement highlighted in current early detection guidance [35].

Pending further understanding of OI‐related HL mechanisms, the novel findings in this paper can be used empirically to update clinical guidance. The OI care team should prioritize early detection of HL in preschool years, establishing a baseline by 9 months of age, in support of early intervention [34]. Hearing monitoring in OI Types III and IV should continue over the lifetime, with more frequent testing for individuals with COL1A1 variants and OI Type III, whereas individuals with COL1A2 may require less frequent monitoring.