Skip to main content
NCBI home page
Communications Biology logoLink to Communications Biology
. 2025 Oct 31;8:1515. doi: 10.1038/s42003-025-08778-2

Large-scale audiometric phenotyping identifies distinct genes and pathways involved in hearing loss subtypes

Samah Ahmed 1, Kenneth I Vaden Jr 2, Morag A Lewis 3, Darren Leitao 4, Karen P Steel 3, Judy R Dubno 2, Britt I Drögemöller 1,5,6,7,✉
PMCID: PMC12579245  PMID: 41174146

Abstract

Age-related hearing loss affects one-third of the population over 65 years. However, the diverse pathologies underlying these heterogeneous phenotypes complicate genetic studies. Here we show that by applying computational phenotyping approaches based on audiometrically measured hearing loss, we can overcome challenges associated with accurate phenotyping for older adults with hearing loss. Using this phenotyping strategy, we uncover differences in the associations observed between genetic variants and sensory and metabolic hearing loss. Sex-stratified analyses of these sexually dimorphic hearing loss phenotypes reveal a locus of relevance to sensory hearing loss in males, but not females. Enrichment analyses implicate genes involved in frontotemporal dementia in metabolic hearing loss, while genes relating to sensory processing of sound by hair cells are implicated in sensory hearing loss. Our study enhances our understanding of these two hearing loss phenotypes, representing the first step in the development of more precise treatments for these pathologically distinct hearing loss phenotypes.

Subject terms: Genetics, Neuroscience

Computational phenotyping of audiometric data reveals differences in the genetic pathways involved in sensory and metabolic hearing loss.

Introduction

Hearing loss affects more than 1.5 billion people around the world, with the costs attributed to unaddressed hearing loss amounting to US$980 billion per year1. In older adults, hearing loss is the most common sensory impairment. In fact, almost one-third of the population over 65 experience difficulties hearing2, with these numbers steadily increasing with an aging population3. The presence of hearing loss is associated with loneliness, stigmatization, depression, and communication difficulties, significantly impacting the quality of life of older individuals4. As age-related hearing loss (ARHL) is a gradually progressive phenotype, early detection before the onset of severe hearing loss is challenging5. For these reasons, improved strategies for the early and precise detection of hearing loss are crucial to guide optimal treatment and management strategies.

ARHL is a multifactorial disorder where genetic predisposition, together with external factors (e.g., noise exposure, aging and certain diseases and ototoxic drugs), contribute to the development of loss of hearing6. In line with this complexity, diverse hearing loss phenotypes involve distinct pathologies7. For example, ARHL that is consistent with excessive exposure to noise, referred to as sensory hearing loss, is typically accompanied by damage to, or death of, cochlear hair cells. These cells are responsible for transforming sound vibrations into electrical signals at specific regions of the cochlea8. In contrast, metabolic hearing loss, is typically associated with deterioration of cells in the stria vascularis, the structure that maintains the endocochlear potential in the inner ear9. These distinct biological underpinnings are further complicated through sex-related differences, with males at higher risk for sensory hearing loss10.

While environmental factors play an important role in ARHL, twin and family-based studies have reported that the heritability of ARHL ranges between 30% and 70%11–14. Genome-wide association studies (GWAS) have uncovered more than 150 loci that are associated with ARHL. Unfortunately, these studies have relied predominantly on self-reported hearing loss15–17. Given the heterogeneity of hearing loss, it is likely that the heritability and genetics underlying the diverse pathologies associated with ARHL will differ. Therefore, understanding the genetic pathways involved in the metabolic and sensory components of ARHL is critical for developing precise and targeted therapies. These differences should also be explored in the context of sex differences, as evidence suggests that sexual dimorphisms related to ARHL may not only reflect differences in environmental risk factors, but also differences in underlying biological processes18.

To address these shortcomings, we applied mathematical approaches to estimate metabolic and sensory components of ARHL using audiogram measurements available through the Canadian Longitudinal Study on Aging (CLSA)19. Using these approaches in combination with comprehensive GWAS approaches, which included the X chromosome and sex-stratified analyses, our study revealed that different genetic variants, genes and cellular mechanisms underlie metabolic and sensory hearing loss, as well as the observed sex-differences in ARHL captured from audiogram-based phenotyping.

Results

Model-based phenotyping to estimate metabolic and sensory hearing loss

Genotype data was available for 26,622 samples included in the CLSA cohort. Following the genotype and phenotype filtering process detailed in Supplementary Figs. 1 and 2, a total of 18,985 samples were selected for further analyses. The mean audiograms of the CLSA participants are shown in Supplementary Fig. 3. In line with previously published work by Vaden et al. 10, the base linear regression model revealed that metabolic hearing loss estimates increased more substantially for older age individuals than sensory estimates (R² = 0.156, P < 0.05 and R² = 0.056, P < 0.05, respectively). This difference was confirmed as statistically significant by a Fisher’s r-to-z comparison test (P < 0.05). Furthermore, one-sided t-tests showed that females had significantly higher metabolic hearing loss estimates than males (P < 0.05), whereas males (vs. females) exhibited significantly higher sensory hearing loss estimates (P < 0.05) (Fig. 1). Analysis of self-reported clinical covariates revealed that older age, sex, diabetes and hypertension were associated with both metabolic and sensory hearing estimates. Osteoporosis and higher sensory hearing estimates were only associated with metabolic hearing estimates. Kidney disease, military service and higher metabolic hearing estimates were only associated with sensory hearing estimates. Variables selected by forward regression analysis are shown in Table 1.

Fig. 1. Comparison of metabolic (A) and sensory (B) hearing loss estimates for the better-hearing ear plotted against age.

Fig. 1

Open in a new tab

Both estimates were calculated from the same participants using the model developed by Vaden et al. The lines represent fitted linear models stratified by sex. A Metabolic estimates increase with age. B Sensory estimates were comparatively less dependent on age and males showed higher sensory hearing loss estimates compared to females. dB: Decibels.

Table 1.

Association between clinical and demographic variables and metabolic and sensory hearing loss

Clinical variable Total cohort Metabolic estimates Sensory estimates
(n = 20,332) P-value Effect size P-value Effect size
Age (years) (mean ± SD) 62.15 ± 10.2 < 2.20×10-16 0.42 < 2.20×10-16 0.38
Metabolic estimates (dB) (mean ± SD) 11.91 ± 12.33 NA NA < 2.20×10-16 -0.60
Sensory estimates (dB) (mean ± SD) 5.62 ± 8.98 < 2.20×10-16 -0.60 NA NA
Sex (male) (n (%)) 9,824 (48.34) < 2.20×10-16 0.01 < 2.20×10-16 0.32
Diabetes (n (%)) 3,449 (16.96) 1.56×10-14 0.07 4.25×10-8 0.05
Osteoporosis (n (%)) 1,712 (8.42) < 2.20×10-16 -0.01 0.97 0.03
Hypertension (n (%)) 7,271 (35.76) < 2.20×10-16 0.03 < 2.20×10-16 0.04
Kidney disease (n (%)) 523 (2.57) 0.02 -0.02 2.28×10-3 0.00
Noisy neighborhood (n (%)) 1,351 (6.64) 0.01 0.00 0.63 0.04
Military service (n (%)) 1,799 (8.85) 0.33 0.00 < 2.20×10-16 -0.01
Open in a new tab

Clinical and demographic covariates associated with each hearing loss estimate are presented with bolded P-values, where significant. Variables selected by forward regression for inclusion in downstream genome-wide association of metabolic estimates are highlighted in blue, while those selected for the sensory estimates are highlighted in red. SD Standard Deviation, dB decibels, NA Not applicable.

Variants and genes associated with metabolic hearing loss

Genome-wide association analyses identified two genomic risk loci that were associated with metabolic hearing loss, but not sensory hearing loss (P < 5 × 10-8; Fig. 2A). Within the first risk locus on chromosome 5, fine-mapping using SuSiE-inf and FINEMAP-inf did not identify any variants in this region with posterior inclusion probability (PIP) > 0.5. However, PolyFun assigned a PIP score nearing this predefined threshold (PIP = 0.48) for rs6453022, the lead variant in this locus (P = 2.67 × 10-9; Supplementary Fig. 4). Annotation analyses revealed that this variant is a missense variant (p.Pro284Gln) in ARHGEF28, a member of the Rho guanine nucleotide exchange factor family16. Investigation of this region revealed that rs6453022 is in high linkage disequilibrium (LD) with several other variants (Supplementary Fig. 5A), likely complicating fine-mapping of this region and leading to the assignment of a relatively low PIP score for the lead variant.

Fig. 2. Association between genetic variants and metabolic and sensory hearing loss.

Fig. 2

Open in a new tab

The results for metabolic and sensory phenotypes are shown in blue and red in the upper and lower panels, respectively. A Miami plot of genome-wide association study analyses. The dashed red line represents genome-wide significance (P < 5 × 10-8). Significant variants are highlighted in green and top variants are labeled. B Miami plot of the gene-based analysis. The red dashed line represents genome-wide significance (2.57 × 10-6). Genes significantly associated with metabolic or sensory estimates are highlighted in green and labeled.

Closer investigation of this region revealed the presence of a second missense variant in ARHGEF28, rs7714670 (p.Trp225Arg) that was also associated with metabolic hearing loss (P = 2.51 × 10-8), which was in high, but incomplete LD, with the lead variant (D’ = 1.0, r2 = 0.85). While these coding variants both result in ARHGEF28 amino acid changes, the impact of the variants on protein function appears to be subtle, with both variants predicted to be benign (SIFT) and tolerated (PolyPhen) for all ARHGEF28 transcripts, with relatively low CADD scores (rs6453022: CADD = 5.42; rs7714670: CADD = 9.01). The occurrence of multiple putative causal variants in this region was further supported by our MAGMA gene-based analyses, which revealed an even stronger association between ARHGEF28 and metabolic hearing loss (MAGMA gene-based P = 4.17 × 10-11 vs. GWAS variant-based P = 2.67 × 10-9). In addition to the association between ARHGEF28 and metabolic hearing loss that was uncovered in the gene-based analyses, a further two autosomal genes were uncovered in these analyses (Fig. 2B) — FUS, located on chromosome 16 (P = 6.17 × 10-7), coding for an RNA binding protein, and IPO7, located on chromosome 11 (P = 1.45 × 10-6), coding for a nuclear import protein.

The second locus that was significantly associated with metabolic hearing loss was uncovered through our XWAS analyses. The lead variant, rs895513076 (P = 1.39 × 10-9; model: complete X-inactivation) with PIP > 0.9, was not predicted to alter the function of any protein coding genes, but was found to occur 2,423 bp upstream of TERF1P7, a processed pseudogene. Although gene-based analysis of the X chromosome did not reveal any genes reaching our predefined genome-wide significance threshold, the association between BCORL1, a transcriptional corepressor, and metabolic hearing loss trended towards significance (P = 3.9 × 10-5). No variants reached genome-wide significance in any of our sex-stratified analysis for the metabolic phenotype.

Variants and genes associated with sensory hearing loss

Our GWAS of sensory hearing loss uncovered a genomic risk locus on chromosome 22 that was significantly associated with sensory hearing loss (P = 2.37 × 10-12; Fig. 2A and Supplementary Fig. 5B), but not metabolic hearing loss (P = 0.001). Fine-mapping of this region revealed that the top variant, rs36062310, had a high PIP score (PIP > 0.9) and is also a missense variant (p.Val504Met; ENST00000395676.4/Val1145Met; ENST00000648057.3) in KLHDC7B, a gene that has been implicated in toxin-mediated ER stress and apoptosis20. This variant has a relatively high CADD score (22.8) and is predicted to have a deleterious (SIFT) and possibly damaging (PolyPhen) effect on the longer protein coding KLHDC7B transcript (ENST00000648057.3).

Gene-based analyses revealed that in addition to the association between KLHDC7B and sensory hearing loss (P = 1.94 × 10-7), a further three autosomal genes were associated with sensory hearing loss (Fig. 2B). This includes two known autosomal recessive non-syndromic hearing loss genes (https://hereditaryhearingloss.org/recessive), CLIC5 on chromosome 6 (P = 4.74 × 10-8) and MYO15A on chromosome 17 (P = 2.38 × 10-7), along with DRG2 (P = 5.46 × 10-7), a GTP-binding protein, located close to MYO15A on chromosome 17. While CLIC5, located on chromosome 6, was associated with sensory hearing loss, no variants or genes in the HLA region reached genome-wide significance for either metabolic or the sensory phenotypes.

Our sex-stratified analyses revealed a previously unreported association between rs72660110 on chromosome 8 and sensory hearing loss. This variant is located proximal to SULF1, a gene that has been shown to play an important role in inner ear development21. This variant was significantly associated with the sensory phenotype in males (P = 2.88 × 10-8), but not females (P = 0.44). A significant interaction was observed between sex and rs72660110 (P = 9.19 × 10-6), suggesting true positive association with the sensory phenotype in males. Fine-mapping of this region revealed that although rs72660110 is the most significant variant, it is not likely to be the causal variant (SuSiE-inf PIP = 0 and FINEMAP-inf PIP = 0.003). While three other variants in this locus were all assigned PIP scores of 1 by both SuSiE-inf and FINEMAP-inf, none of these variants were identified as likely causal by PolyFun (PIP < 0.5), with rs72660104, an intronic variant, receiving the highest score in this region (PIP = 0.289). Additionally, we uncovered that rs36062310 (p.Val1145Met; KLHDC7B), the sensory hearing loss-associated variant, was more significantly associated with this phenotype in females compared to males (combined P = 2.37 × 10-12; females P = 8.01 × 10-9; males P = 4.00 × 10-5). However, the interaction between sex and rs36062310 was not significant (P = 0.62), indicating no evidence of a sex-dependent effect for this variant. A list of the potential causal variants is shown in Table 2.

Table 2.

Top genomic risk loci significantly associated with metabolic and sensory hearing loss phenotypes

Lead SNP Metabolic (P-value) Sensory (P-value) Variant consequence Nearest gene FINEMAP-inf score (top variant) SuSiE-inf score (top variant) PolyFun score (top variant) Implicated in Hearing Loss
Combined Males Females Combined Males Females
rs6453022 3×10-9 7×10-4 5×10-7 5×10-4 0.14 4×10-4 Missense (Pro284Gln) ARHGEF28 0.01 (rs4413512) 0.05 (rs79391401) 0.48 (rs6453022) Yes, ARHL11
rs36062310 0.001 0.23 5×10-4 2×10-12 4×10-5 8×10-9 Missense (Val504Met/ Val1145Met) KLHDC7B 0.99 (rs36062310) 0.99 (rs36062310) 0.99998 (rs36062310) Yes, ARHL11
rs72660110 0.003 8×10-5 0.80 4×10-4 3×10-8 0.44 Downstream of gene SULF1 1.00 (rs4279601) 1.00 (rs4279601) 0.289 (rs72660104) Yes, inner ear development21
1.00 (rs10957501) 1.00 (rs10957501)
1.00 (rs11785839) 1.00 (rs11785839)
1.00 (rs62512191) 0.93 (rs62512191)
rs895513076 1×10-9 4×10-4 1×10-7 1×10-3 0.12 8×10-4 Upstream of pseudogene TERF1P7 0.96 (rs895513076) 0.84 (rs895513076) NA No
Open in a new tab

P-values shaded in blue represent variants significantly association with metabolic estimates, while those shaded in red represent variants significantly associated with the sensory estimates. Significance is determined at the genome-wide threshold of 5 × 10-8. ARHL Age-related hearing loss, SNP Single nucleotide polymorphism.

Comparison of metabolic and sensory associations

We compared the standardized effect sizes of the genetic variants that were identified by our PolyFun analyses for both hearing loss phenotypes. The effect sizes of variants uncovered from both GWAS were not significantly correlated with each other, with metabolic-associated variants showing a negligible inverse correlation with their effects in the sensory phenotype (r = –0.03), and sensory-associated variants demonstrating a moderate inverse correlation with metabolic effects (r = –0.35) (Fig. 3). In line with this, examination of the genetic correlation between metabolic and sensory hearing loss using Linkage Disequilibrium Score Regression (LDSC) analyses revealed a significant negative genetic correlation (rg = -0.30, se=0.11, P = 0.007), consistent with the observed inverse phenotypic correlation (Table 1).

Fig. 3. Comparison of the standardized effect sizes of variants associated with metabolic and sensory hearing loss phenotypes.

Fig. 3

Open in a new tab

Effect sizes of variants that were significantly associated with metabolic estimates (shown in blue) and sensory estimates (shown in red) do not show the same direction of effect for both hearing loss phenotypes (slope: -0.01; 95% CI: -0.24-0.22; P = 0.92 and slope: -2.24; 95% CI: -7.61-3.12; P = 0.36, respectively).

Replication analyses

Our GWAS identified three loci that were significantly associated with metabolic and sensory hearing loss (Fig. 2). We were able to extract coding variants located within two of these loci (the metabolic hearing loss risk locus on chromosome 5 and the sensory hearing loss risk locus on chromosome 22) for investigation in the replication cohort. For the chromosome 5 locus, we were able to extract information for two missense variants in ARHGEF28 (rs6453022; p.Pro284Gln and rs7714670; p.Trp225Arg). Investigation of these variants in the replication cohort revealed that they were both associated with the metabolic phenotype (P = 0.01 and P = 0.03, respectively), but not the sensory phenotype (P = 0.62 and P = 0.80, respectively). Investigation of the missense variant located on chromosome 22 (rs36062310; p.Val504Met) did not uncover significant associations with either the sensory or metabolic phenotypes (P = 0.99 and P = 0.10, respectively).

Enrichment of genetic variants in Mendelian hearing loss genes

Investigation of the GWAS results revealed that half of the genes that were significantly associated with sensory hearing loss have previously been reported to cause autosomal recessive non-syndromic hearing loss. To investigate this further, we used a quantile-quantile (QQ) plot to visualize observed vs. expected P values for variants occurring in Mendelian hearing loss genes in comparison to other genes (Fig. 4). These analyses revealed that while variants in Mendelian hearing loss genes were enriched in both hearing loss phenotypes, this effect was more pronounced for the sensory hearing loss phenotype.

Fig. 4. QQ plots of observed versus expected -log10 (P-values) for variants associated with sensory and metabolic hearing loss phenotypes.

Fig. 4

Open in a new tab

The left plot shows results for the metabolic phenotype, and the right plot shows results for the sensory phenotype. Blue and red dots represent variants within all genes (n = 19,438, including the X chromosome) for the metabolic and sensory phenotype, respectively, and gray dots represent variants within Mendelian hearing loss genes (n = 193).

To investigate whether variants in the same Mendelian deafness genes were driving the observed enrichments, we extracted all variants occurring within Mendelian hearing loss genes that were associated (P < 1 x 10-3 i.e., P < 0.05 after correction for multiple testing of 193 Mendelian hearing loss genes) with either the metabolic or sensory hearing loss phenotypes. While variants in certain Mendelian hearing loss genes were associated with both phenotypes (PCDH15, TSPEAR, ESPN, GPR98 HOMER2 and REST), most variants in these Mendelian hearing loss genes were only associated with either the metabolic (n = 29 genes) or sensory (n = 19 genes) phenotype.

Pathway analyses with EnrichR revealed that Mendelian hearing loss genes that were associated with the sensory phenotype were more likely to be implicated in pathways involved in the sensory processing of sound (P = 1.51 × 10-10, OR = 113.85), potassium cycling in noise-induced hearing loss (7.403 × 10-10, OR = 311.86), gap junction trafficking (P = 8.28 × 10-5, OR = 123.68), and hair cell stereocilia protein dysfunction resulting in both congenital (P = 4.95 × 10-18; OR = 415.58) and age-related (P = 6.03 × 10-13; OR = 573.13) hearing loss. Pathways that were more significantly enriched in the metabolic phenotype included collagen synthesis (P = 0.002, OR = 45.56), and processes related to general tissue maintenance and structural integrity. A detailed list of variants and mapped genes is provided in Supplementary Data 1.

Discussion

By applying an approach developed by Vaden et al.10 to calculate metabolic and sensory hearing loss estimates from individual audiograms we were able to uncover unique genetic pathways underlying two distinct hearing loss phenotypes. The findings delineated the role of genes that have previously been associated with self-reported hearing loss in specific hearing loss phenotypes. Further, through the inclusion of specialized X-chromosome, gene-based and sex-stratified analyses, we were able to uncover distinct associations between biologically plausible genes and metabolic and sensory components of ARHL, including some sex-specific associations.

Our GWAS analyses revealed that two missense variants, rs6453022 (p.Pro284Gln) and rs7714670 (p.Trp225Arg), in ARHGEF28 were associated with the metabolic hearing loss phenotype. Although these variants have previously been reported to be associated with self-reported hearing difficulty (Supplementary Data 2)15, to the best of our knowledge, this is the first study to define their specific role in metabolic hearing loss. Notably, these findings were replicated in an independent cohort, further supporting their role in the metabolic, but not sensory, component of ARHL. While the precise mechanism through which ARHGEF28 contributes to ARHL remains unclear, previous research has indicated its potential involvement in the regulation of neurofilaments, as well as axon growth and branching22. In addition to its expression in the stria vascularis16, ARHGEF28 is expressed in both inner ear hair cells and spiral ganglion neurons15. Therefore, dysregulation of this gene could disrupt the transmission of electrical signals to the central auditory system and the brain. Interestingly, ARHGEF28 variants have also been associated with dementia-related TD-43 pathology23, and over-expression of the N-terminal fragment of this protein has been shown to reduce neurodegeneration and neuroinflammation in fly and mouse models24. In line with this, our pathway analyses revealed that genes associated with metabolic hearing loss were also implicated in amyotrophic lateral sclerosis and fronto-temporal dementia23,25,26. Given that hearing loss has been reported to be one of the largest modifiable risk factors for dementia27–29, future studies are warranted to determine whether this relationship may, in part, be attributed to dysregulation of shared genetic pathways.

Investigation of sensory hearing loss also uncovered a significant association between a missense variant, rs36062310 (p.Val504Met), in KLHDC7B, which has also been strongly associated with self-reported hearing difficulty in previous studies15–17,30. Further investigation of this variant in the replication cohort did not uncover a significant association with either the sensory or metabolic phenotype. This may be attributed to the low allele frequency of this variant in the replication cohort (MAF = 0.04, with only one homozygous individual). As this variant has previously been associated with self-reported hearing difficulty, the findings uncovered in the discovery cohort warrant further investigation in a larger cohort of carefully phenotyped individuals. While the exact mechanism by which KLHDC7B influences ARHL is still under investigation, ongoing research suggests that it plays a crucial role in toxin mediated apoptosis and maintaining cochlear hair cells20,31.

A prominent theory regarding the genetics underlying ARHL points to the involvement of genetic variants in Mendelian hearing loss genes32. In line with this, our gene-based analysis identified MYO15A and CLIC5 as significant contributors to sensory hearing loss. These genes are implicated in autosomal recessive non-syndromic hearing loss (DFNB3 and DFNB103, respectively)33,34. Broader investigation of the importance of Mendelian hearing loss genes to ARHL revealed that variants occurring within Mendelian hearing loss genes were more likely to be associated with both sensory and metabolic hearing loss when compared to other genes (Fig. 4). With specific reference to sensory hearing loss, these Mendelian hearing loss genes were enriched in pathways related to sensory processing of sound by inner and outer hair cells. This is in alignment with reports that sensory hearing loss is typically accompanied by damage to, or death of, cochlear hair cells8.

Sex-stratified GWAS analyses revealed key differences for sensory hearing loss, which is a sexually dimorphic trait35. Firstly, these analyses revealed an association between rs36062310 (p.Val1145Met; KLHDC7B) and sensory hearing loss. Although the sex-genotype interaction was not statistically significant, supporting evidence from the International Mouse Phenotyping Consortium (IMPC) (https://www.mousephenotype.org/) shows abnormal hearing morphology in mice with Klhdc7b mutations in combined (male and female) and females samples, but not in males alone. Further, these analyses uncovered a previously unreported association between rs72660110, a variant mapping to SULF1, and sensory hearing loss in males only. Although this gene has not been previously implicated in ARHL, it has been shown to play an important role in the development of the inner ear in animal models21. SULF1 is also one of three genes that were reported to show major differences in gene expression in aged female mice compared to aged male mice36. Although the exact consequences of the variants in this locus remain to be elucidated, these findings have revealed the importance of genetics in the observed sex differences for sensory hearing loss.

Previous studies have reported that the role of the X-chromosome in ARHL is limited37. However, by including the X-chromosome in our analysis of hearing sub-types, we have identified a locus on the X chromosome that was associated with the metabolic phenotype. Even though annotation of this region was complicated by limited annotation information for variants occurring on the X chromosome, gene-based analyses uncovered a trend towards significance between BCORL1 and metabolic hearing loss. This gene has previously been implicated in intellectual disability with hearing loss, and has been shown to contribute to age-related epigenetic alterations38–40. Therefore, this previously unreported association warrants further investigation.

In addition to identifying different risk loci for metabolic and sensory hearing loss, our results also revealed that the effect sizes of these variants were not significantly correlated across the phenotypes, with a trend towards negative correlation observed (Fig. 3), suggesting that different genetic pathways are involved in these two phenotypes. This is supported by the significant negative genetic correlation that was observed between metabolic and sensory hearing loss. Indeed, while most older adults demonstrate a combination of metabolic and sensory hearing loss10,41,42, suggesting these traits may influence one another, a negative phenotypic correlation was also observed between metabolic and sensory hearing loss (Table 1).

While our study provided valuable insights into the genetic pathways underlying different ARHL phenotypes, we also acknowledge several limitations. First, examination of audiograms revealed that hearing measurements taken at 0.5 kHz appear slightly elevated due to background noise during audiometric testing, as previously reported43. While this limitation should be acknowledged, the overall configurations of these audiograms remain consistent with the phenotyping model that was applied to this dataset. Second, the CLSA database relies on self-reported phenotypes for many variables and contains limited information for certain demographic/clinical variables such as occupational and recreational noise exposure – key factors that influence the progression of ARHL, especially with regards to the sensory phenotype. To help disentangle the effects of such unmeasured environmental exposures, future studies could benefit from applying approaches such as Mendelian Randomization44. Lastly, our cohort is predominantly composed of individuals of European ancestry ( > 90%), with other populations under-represented, which may impact the generalizability of our findings to more diverse groups. This limitation is particularly relevant given that the ARHGEF28 and KLHDC7B variants uncovered in this study are most common in individuals of European ancestry. For example, rs36062310 – the missense variant in KLHDC7B that was significantly associated with sensory hearing loss in our study – has a MAF of 0.04 in individuals of European ancestry, and is extremely rare in individuals of African and East Asian ancestry (0.007 and 9×10⁻⁵, respectively)45. Moreover, variants of greater relevance to non-European populations may be entirely missed in analyses restricted to European-ancestry cohorts, highlighting the need for future studies in more diverse populations.

Conclusion

Our study has confirmed the heterogeneous nature of ARHL and, to the best of our knowledge, is the first to use GWAS to uncover specific genetic pathways that are of relevance to distinct hearing loss phenotypes. The identification of specific genes and pathways underlying metabolic and sensory hearing loss holds significant promise for the identification of drug targets for each hearing loss phenotype. For instance, genes such as ARHGEF28 and KLHDC7B, which we identified as important contributors to metabolic and sensory hearing loss, respectively, exhibit high genetic priority scores (2.8 and 2.1, respectively). This is important as genes with Genetics-guided Priority Score (GPS) > 2.1 have been shown to have a more than 10-fold higher chance of success in Phase IV drug development trials46. These two genes are also implicated in critical biological processes related to hearing. Therefore, if additional evidence for the role that these genes and variants play in ARHL is provided through functional validation using animal models, they may provide potential targets for future therapeutic exploration in the treatment of ARHL. Taken together, this study has improved our understanding of the genetics underlying sensory and metabolic hearing loss and opened new avenues for future research aimed at improving early diagnosis and precise treatment of hearing loss in older adults.

Methods

Patient cohort and phenotyping of hearing loss

We obtained genotype and baseline audiogram, demographic and clinical data from 30,097 individuals included in the CLSA47. Hearing thresholds were measured without the use of hearing aids at 0.5, 1, 2, 3, 4, 6 and 8 kHz by pure-tone signals ranging from 0 to 100 dB HL, with 5 dB increments, using a digital screening audiometer48. Individuals with more than one missing audiogram measurement or unreliable audiometric tests were excluded from downstream analyses. Additionally, individuals were excluded if they exhibited conditions indicative of conductive hearing loss, such as the presence of ear wax, collapsed ear canals, or ear infections (Supplementary Fig. 1). Audiometric phenotyping was determined by adopting a mathematical modeling approach developed by Vaden et al.10. This approach involved fitting individual audiogram data to previously defined hearing loss templates, which were based on average audiograms considered to be exemplars of metabolic and sensory hearing loss42. In this way, we were able to obtain estimates for the extent of metabolic and sensory hearing loss for each ear in each participant. Audiograms that are well-approximated by a combination of metabolic and sensory templates had line fit error values < 15 dB (i.e., low predicted error), and were selected for inclusion in downstream analyses. Further, for each individual, we identified ears with better and worse hearing based on the calculated sensory and metabolic estimates. Age-related trends in hearing loss estimates were assessed using linear regression. To compare the strength of association between age and the metabolic versus sensory estimates, correlations were statistically compared using Fisher’s r-to-z transformation for dependent overlapping correlations, implemented with the R package cocor (version 1.1.4). Sex differences were evaluated with one-sided t-tests. Statistical significance was set at P < 0.05. This study was approved by the local ethics committee.

Genetic data and heritability analyses

Genotype data were previously generated using the Affymetrix Axiom array and all samples underwent quality control (QC) and imputation using the TOPMed reference panel as previously described48. Samples with call rates ≥ 95% were included in the analyses and related individuals (pairwise kinship coefficient > 0.125), as well as those with extreme heterozygosity, genotype missingness, and discordant genetic and self-reported sex, were excluded as previously described48. For marker-based QC, variants were removed based on the following criteria: variant call rates≤95%, imputation quality (R2) < 0.8, minor allele frequency (MAF) < 0.05, and Hardy-Weinberg Equilibrium (HWE) P < 1×10-6 (Supplementary Fig. 2). Samples and markers that passed genotype and phenotype QC were included to estimate SNP-based heritability for better and worse hearing ears using GCTA-GREML49. For both metabolic and sensory estimates, the ear of each participant with the better hearing showed higher heritability (h2 = 0.08, P = 9.03 × 10-7 and h2 = 0.11, P = 1.37 × 10-11, respectively), compared to the ear with worse hearing (h2 = 0.058, P = 4.49 × 10-4 and h2 = 0.105, P = 1.19×10-10, respectively). Therefore, the better ear for each participant was used in all downstream genetic analyses. Better-ear metabolic and sensory estimates were normalized using the rank function in the base R Statistics software (v4.3.2)50.

Clinical and genome-wide association analyses

Linear regression was used to identify demographic and clinical variables that were associated with sensory and metabolic estimates. To minimize the impact of mixed phenotypes, we included the alternate phenotype as a covariate in our GWAS models. Forward regression was subsequently performed on significant variables (P < 0.05, after correction for multiple testing). Variables uncovered from these analyses, along with the first ten genetic principal components provided by the CLSA48, were included as covariates in the downstream genomic analyses. For autosomal chromosomes, we performed two separate GWAS using linear regression to identify genetic variants that were associated with sensory or metabolic estimates using PLINK (v1.9 and v2.0)51,52. To examine polymorphisms in the Human Leukocyte Antigen (HLA) region, we converted CLSA imputed HLA alleles to PLINK format using the HLA Analysis Toolkit (HATK; v2.0) and included these alleles in linear regression analyses53. As hearing loss has been shown to be sexually dimorphic54, we conducted sex-stratified association testing for each hearing loss estimate. Sex×genotype interaction analyses were performed for lead significant variants to assess the presence of sex-specific effects, with significance defined at P < 0.05. Finally, we examined the association between genetic variants on the X chromosome and metabolic and sensory estimates using chromosome X-Wide Analysis toolSet (XWAS)55. This tool applies sex-stratified QC steps, i.e., missingness, MAF and HWE (Supplementary Fig. 2) to account for differences in variant calling in males and females. We conducted the XWAS analyses using two models. The first model assumed complete X-inactivation, where hemizygous males were coded as having 0 or 2 alleles, corresponding to homozygous females. The second model assumed escape from X-inactivation, where males were coded as having 0 and 1 alleles corresponding to heterozygous females. Bonferroni-corrected threshold of P < 5×10-8 was considered genome-wide significant in all SNP-based analyses. Miami plots were generated using ‘miamiplot v1.1.0’ R package.

Fine-mapping analyses

To uncover likely causal variants within each of the identified genomic risk loci, we defined each region by taking the lead variant flanked by 50,000 base pairs. We then implemented statistical fine-mapping analyses using Sum of Single Effects-inf (SuSiE-inf) and FINEMAP-inf methods to prioritize the variants which are most likely to be causal within each region56. Variants with PIP > 0.5 from both methods were considered likely causal. These fine-mapping analyses were further enhanced by including functional annotation using POLYgenic FUNctionally informed fine-mapping (PolyFun). Variants meeting the suggestive level of significance (P < 1×10-5) were included in these analyses57. Variants uncovered from these analyses were annotated using Ensembl’s Variant Effect Predictor (VEP)58 and the Combined Annotation Dependent Depletion (CADD) scoring system59, using default parameters.

Gene and enrichment analyses

To examine the combined effects of variants within a genomic region, we performed gene-based association testing for the autosomal chromosomes using Multi-marker Analysis of GenoMic Annotation (MAGMA), implemented through FUMA60. For the X-chromosome, gene-based association analyses were performed using the XWAS tool, using a modified versatile gene-based association study (VEGAS) framework55,61. Genes that reached a Bonferroni-corrected significance threshold (P < 2.57×10-6 i.e., adjusted for 19,438 genes) were considered significant.

Several Mendelian hearing loss genes were found to be associated with the hearing loss phenotypes included in our analyses. We therefore compared the enrichment of GWAS variants located within 50 kb of Mendelian hearing loss genes (recorded in the Hereditary Hearing Loss Database https://hereditaryhearingloss.org/)16,17 and non-Mendelian genes. This comparison aimed to investigate whether these hearing loss variants were more likely to be associated with sensory or metabolic hearing loss. Pathway analyses were performed using ErichR62 to identify specific processes related to the Mendelian hearing loss genes associated with sensory or metabolic estimates. Pathways with a Benjamini-Hochberg adjusted P < 0.05 were considered significant.

Genetic relationship between metabolic and sensory phenotypes

To investigate the genetic relationship between the two hearing loss phenotypes, we conducted GWAS for each phenotype in participants of European ancestry. In the primary analyses, each phenotype was adjusted for the other (in addition to standard covariates in Table 1) to isolate SNP effects specific to each trait. However, to compare effect sizes and estimate genetic correlation across phenotypes, we also generated GWAS summary statistics without including the opposite phenotype as a covariate. Using these unadjusted GWAS results, we compared the effects of variants identified through PolyFun fine-mapping by fitting a linear regression model between the standardized effect sizes for the metabolic and sensory phenotypes.

To estimate genetic correlation between metabolic and sensory hearing loss, we used LDSC. Summary statistics from the unadjusted GWAS model were used as input for these analyses. These summary statistics were aligned with HapMap3 variants, and the 1000 Genomes Phase 3 European LD panel was used as a reference. A total of 987,960 variants from our summary statistics overlapped with the LD reference panel, indicating good concordance between the datasets.

Replication analyses

To replicate significant findings from the discovery cohort, a replication cohort consisting of 502 participants aged 55 years and older from the Medical University of South Carolina (MUSC) Longitudinal Cohort Study of ARHL was used. All participants in the replication cohort were of non-Finnish European ancestry. The MUSC and CLSA cohorts were broadly similar in terms of age distribution and hearing thresholds, allowing for meaningful comparison and replication of findings42. Whole-exome sequencing, variant calling, and QC procedures for this cohort have been previously described by Lewis et al.7. Metabolic and sensory components of ARHL were derived from audiometric data using the same approach outlined by Vaden et al.10. As only exome sequencing data were available for this cohort, variants within the coding regions of the genome were extracted for replication purposes. A linear regression model was used to examine associations between coding variants uncovered in the discovery cohort and rank-transformed metabolic and sensory phenotypes, including the same covariates used in the discovery analyses (age, sex where applicable, alternate hearing phenotype, and the first 10 PCs). Information on diabetes and hypertension were not available in the replication dataset and were therefore not included as covariates. Sex-stratified analyses were not performed due to small sample sizes.

Statistics and reproducibility

Statistical analyses were performed using R (v4.3.2). GWAS were conducted using PLINK (v1.9 and v2.0) and XWAS following quality control procedures (Supplementary Fig. 2). Gene-based analyses were performed using MAGMA (v1.6.0) via FUMA GWAS, and fine-mapping analyses were conducted using SuSiE-inf (v1.2), FINEMAP-inf (v1.3), and PolyFun (FINEMAP v1.4.1). LD score regression was performed with LDSC (v1.0.1). All analyses were conducted using default parameters unless otherwise specified. The discovery cohort comprised n = 26,622 participants from the CLSA who met inclusion criteria and passed quality control (Supplementary Fig. 1). Replication analyses were conducted in an independent cohort of 502 participants.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information (929.6KB, pdf)
42003_2025_8778_MOESM2_ESM.pdf (88.2KB, pdf)

Description of Additional Supplementary Materials

Supplementary Data 1 (55.8KB, xlsx)
Supplementary Data 2 (18.2KB, xlsx)
Reporting summary (2.7MB, pdf)
Transparent Peer Review file (472.1KB, pdf)

Acknowledgements

This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). Funding for the CLSA is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation, as well as the following provinces, Newfoundland, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia. This research has been conducted using the CLSA Baseline Comprehensive Dataset Version 6.0 and Genome-wide Genetic Data Version 3.0 under Application Number 2104035. The CLSA is led by Drs. Parminder Raina, Christina Wolfson and Susan Kirkland. The opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging. This work was funded through a Natural Sciences and Engineering Research Council of Canada Discovery Grant (B.I.D.) and (in part) by the National Institutes of Health/National Institute on Deafness and Other Communication Disorders Clinical Research Center (P50 DC 000422) awarded to the Medical University of South Carolina and by the South Carolina Clinical and Translational Research (SCTR) Institute, with an academic home at the Medical University of South Carolina, NIH/NCATS Grant number UL1 TR001450 (J.R.D. and K.V.). Portions of this investigation were conducted in a facility constructed with support from Research Facilities Improvement Program Grant Number C06 RR14516 from the NIH/NCRR (J.R.D. and K.V.). B.I.D. is supported by a CIHR Tier 2 Canada Research Chair in Pharmacogenomics and Precision Medicine. S.A. was supported through a Research Manitoba Studentship and the NSERC-CREATE Visual and Automated Disease Analytics Graduate Training Program at the University of Manitoba.

Author contributions

S.A. performed the discovery analyses and wrote the manuscript. K.I.V. and J.R.D. developed the phenotyping approach and provided clinical expertise. D.L. provided clinical expertise. M.A.L. and K.P.S. performed replication analysis. B.I.D. conceived of and supervised the project. All authors contributed to editing the manuscript.

Peer review

Peer review information

Communications Biology thanks Yongyi Yuan and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Huiying Zhao and Aylin Bircan, David Favero. A peer review file is available.

Data availability

Individual level data are available from the Canadian Longitudinal Study on Aging. Interested researchers should submit an online application including a research proposal. Researchers who meet the access criteria will receive de-identified data via secure download. Full instructions are available at https://www.clsa-elcv.ca/data-access/. GWAS and XWAS summary statistics generated in this study are publicly available via Zenodo at 10.5281/zenodo.16897822. All other data supporting the findings of this study are available in Supplementary Data 1 and 2.

Code availability

Scripts that were used in this study are available on https://github.com/Drogemoller-Lab/ARHL.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-025-08778-2.

References

  • 1.Deafness and hearing loss. https://www.who.int/news-room/fact-sheets/detail/deafness-and-hearing-loss.
  • 2.Wattamwar, K. et al. Increases in the rate of age-related hearing loss in the older old. JAMA Otolaryngol. Head. Neck Surg.143, 41–45 (2017). [DOI] [PubMed] [Google Scholar]
  • 3.Prince, M. J. et al. The burden of disease in older people and implications for health policy and practice. Lancet385, 549–562 (2015). [DOI] [PubMed] [Google Scholar]
  • 4.Kotby, M. N., Tawfik, S., Aziz, A. & Taha, H. Public health impact of hearing impairment and disability. Folia Phoniatr. Logop.60, 58–63 (2008). [DOI] [PubMed] [Google Scholar]
  • 5.Tsimpida, D., Kontopantelis, E., Ashcroft, D. & Panagioti, M. Comparison of self-reported measures of hearing with an objective audiometric measure in adults in the english longitudinal study of ageing. JAMA Netw. Open3, e2015009 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Yamasoba, T. et al. Current concepts in age-related hearing loss: epidemiology and mechanistic pathways. Hear Res.303, 30–38 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Lewis, M. A. et al. Accurate phenotypic classification and exome sequencing allow identification of novel genes and variants associated with adult-onset hearing loss. PLoS Genet.19, e1011058 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Wu, O’Malley, P., Gruttola, J. T., de, V. & Liberman, M. C. Age-related hearing loss is dominated by damage to inner ear sensory cells, not the cellular battery that powers them. J. Neurosci.40, 6357–6366 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Schmiedt, R. A. Effects of aging on potassium homeostasis and the endocochlear potential in the gerbil cochlea. Hear Res. 102, 125–132 (1996). [DOI] [PubMed] [Google Scholar]
  • 10.Vaden, K. I., Eckert, M. A., Matthews, L. J., Schmiedt, R. A. & Dubno, J. R. Metabolic and sensory components of age-related hearing loss. JARO23, 253–272 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Karlsson, K. K., Harris, J. R. & Svartengren, M. Description and primary results from an audiometric study of male twins. Ear Hear18, 114–120 (1997). [DOI] [PubMed] [Google Scholar]
  • 12.Gates, G. A., Couropmitree, N. N. & Myers, R. H. Genetic associations in age-related hearing thresholds. Arch. Otolaryngol. Head. Neck Surg.125, 654–659 (1999). [DOI] [PubMed] [Google Scholar]
  • 13.Wingfield, A. et al. A Twin-Study of Genetic Contributions to Hearing Acuity in Late Middle Age. J. Gerontol. A Biol. Sci. Med Sci.62, 1294–1299 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Duan, H. et al. Heritability of Age-Related Hearing Loss in Middle-Aged and Elderly Chinese: A Population-Based Twin Study. Ear Hearing40, 253 (2019). [DOI] [PubMed] [Google Scholar]
  • 15.Wells, H. R. R. et al. GWAS Identifies 44 Independent Associated Genomic Loci for Self-Reported Adult Hearing Difficulty in UK Biobank. Am. J. Hum. Genet105, 788–802 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Kalra, G. et al. Biological insights from multi-omic analysis of 31 genomic risk loci for adult hearing difficulty. PLOS Genet.16, e1009025 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Liu, W., Johansson, Å, Rask-Andersen, H. & Rask-Andersen, M. A combined genome-wide association and molecular study of age-related hearing loss in H. sapiens. BMC Med.19, 302 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Nolan, L. S. Age-related hearing loss: Why we need to think about sex as a biological variable. J. Neurosci. Res.98, 1705–1720 (2020). [DOI] [PubMed] [Google Scholar]
  • 19.Mick, P. T. et al. The Prevalence of Hearing, Vision, and Dual Sensory Loss in Older Canadians: An Analysis of Data from the Canadian Longitudinal Study on Aging. Can. J. Aging40, 1–22 (2021). [DOI] [PubMed] [Google Scholar]
  • 20.Yahiro, K. et al. A novel endoplasmic stress mediator, Kelch domain containing 7B (KLHDC7B), increased Harakiri (HRK) in the SubAB-induced apoptosis signaling pathway. Cell Death Discov.7, 1–11 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Freeman, S. D., Keino-Masu, K., Masu, M. & Ladher, R. K. Expression of the heparan sulfate 6-O-endosulfatases, Sulf1 and Sulf2, in the avian and mammalian inner ear suggests a role for sulfation during inner ear development. Dev. Dyn.244, 168–180 (2015). [DOI] [PubMed] [Google Scholar]
  • 22.Volkening, K., Leystra-Lantz, C. & Strong, M. J. Human low molecular weight neurofilament (NFL) mRNA interacts with a predicted p190RhoGEF homologue (RGNEF) in humans. Amyotroph. Lateral Scler.11, 97–103 (2010). [DOI] [PubMed] [Google Scholar]
  • 23.Katsumata, Y. et al. Genetic associations with dementia-related proteinopathy: Application of item response theory. Alzheimers Dement20, 2906–2921 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Droppelmann, C. A. et al. Mitigation of TDP-43 toxic phenotype by an RGNEF fragment in amyotrophic lateral sclerosis models. Brain147, 2053–2068 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Neumann, M. et al. A new subtype of frontotemporal lobar degeneration with FUS pathology. Brain132, 2922–2931 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Kim, W., Kim, D.-Y. & Lee, K.-H. RNA-Binding Proteins and the Complex Pathophysiology of ALS. Int J. Mol. Sci.22, 2598 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Son, S. et al. Potentially modifiable dementia risk factors in Canada: an analysis of canadian longitudinal study on aging with a multi-country comparison. J. Prev. Alzheimers Dis.10.14283/jpad.2024.105 (2024). [DOI] [PMC free article] [PubMed]
  • 28.Livingston, G. et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet404, 572–628 (2024). [DOI] [PubMed] [Google Scholar]
  • 29.Uchida, Y. et al. Age-related hearing loss and cognitive decline — The potential mechanisms linking the two. Auris Nasus Larynx46, 1–9 (2019). [DOI] [PubMed] [Google Scholar]
  • 30.Trpchevska, N. et al. Genome-wide association meta-analysis identifies 48 risk variants and highlights the role of the stria vascularis in hearing loss. Am. J. Hum. Genet.109, 1077–1091 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Kaufman, A. Exploring the Function of KLHDC7b, a Novel Gene Associated With Hearing Loss. in (Anaheim, CA, USA, 2024).
  • 32.Ivarsdottir, E. V. et al. The genetic architecture of age-related hearing impairment revealed by genome-wide association analysis. Commun. Biol.4, 1–13 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Wang, A. et al. Association of unconventional myosin MYO15 mutations with human nonsyndromic deafness DFNB3. Science280, 1447–1451 (1998). [DOI] [PubMed] [Google Scholar]
  • 34.Wonkam-Tingang, E. et al. Bi-allelic novel variants in CLIC5 identified in a cameroonian multiplex family with non-syndromic hearing impairment. Genes (Basel)11, 1249 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Nam, D. W. et al. Sex differences in associated factors for age-related hearing loss. PLoS ONE19, e0298048 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Wang, Y. et al. Sex differences in transcriptomic profiles in aged kidney cells of renin lineage. Aging (Albany NY)10, 606–621 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Naderi, E. et al. The genetic contribution of the X chromosome in age-related hearing loss. Front. Genet. 14, (2023). [DOI] [PMC free article] [PubMed]
  • 38.Barone, C. et al. Intragenic ILRAPL1 deletion in a male patient with intellectual disability, mild dysmorphic signs, deafness, and behavioral problems. Am. J. Med. Genet. A161A, 1381–1385 (2013). [DOI] [PubMed] [Google Scholar]
  • 39.Shukla, A. et al. Variants in the transcriptional corepressor BCORL1 are associated with an X-linked disorder of intellectual disability, dysmorphic features, and behavioral abnormalities. Am. J. Med. Genet. A179, 870–874 (2019). [DOI] [PubMed] [Google Scholar]
  • 40.Wonkam, A. et al. Exome sequencing of families from Ghana reveals known and candidate hearing impairment genes. Commun. Biol.5, 369 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Schmiedt, R. A. The physiology of cochlear presbycusis. in The Aging Auditory System (eds. Gordon-Salant, S., Frisina, R. D., Popper, A. N. & Fay, R. R.) 9–38 (Springer, New York, NY, 2010). 10.1007/978-1-4419-0993-0_2.
  • 42.Dubno, J. R., Eckert, M. A., Lee, F.-S., Matthews, L. J. & Schmiedt, R. A. Classifying human audiometric phenotypes of age-related hearing loss from animal models. J. Assoc. Res. Otolaryngol.14, 687–701 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Mick, P. T. et al. Associations between cardiovascular risk factors and audiometric hearing: findings from the Canadian longitudinal study on aging. Ear Hearing44, 1332 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Sanderson, E. et al. Mendelian randomization. Nat. Rev. Methods Prim.2, 1–21 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Karczewski, K. J. et al. The mutational constraint spectrum quantified from variation in 141,456 humans. Nature581, 434–443 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Duffy, Á. et al. Development of a human genetics-guided priority score for 19,365 genes and 399 drug indications. Nat. Genet.56, 51–59 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Raina, P. et al. Cohort Profile: The Canadian Longitudinal Study on Aging (CLSA). Int. J. Epidemiol.48, 1752–1753j (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Forgetta, V. et al. Cohort profile: genomic data for 26 622 individuals from the Canadian Longitudinal Study on Aging (CLSA). BMJ Open12, e059021 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Yang, J., Lee, S. H., Goddard, M. E. & Visscher, P. M. GCTA: a tool for genome-wide complex trait analysis. Am. J. Hum. Genet.88, 76–82 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.R Core Team. R: A Language and Environment for Statistical Computing_. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2023).
  • 51.Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet81, 559–575 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Purcell, S. PLINK v1.09, v2.0. http://pngu.mgh.harvard.edu/purcell/plink/.
  • 53.Choi, W., Luo, Y., Raychaudhuri, S. & Han, B. HATK: HLA analysis toolkit. Bioinformatics37, 416–418 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.De Angelis, F. et al. Sex differences in the polygenic architecture of hearing problems in adults. Genome Med.15, 1–18 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Gao, F. et al. XWAS: a software toolset for genetic data analysis and association studies of the X chromosome. J. Hered.106, 666–671 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Cui, R. et al. Improving fine-mapping by modeling infinitesimal effects. Nat. Genet.56, 162–169 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Weissbrod, O. et al. Functionally informed fine-mapping and polygenic localization of complex trait heritability. Nat. Genet.52, 1355–1363 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.McLaren, W. et al. The Ensembl Variant Effect Predictor. Genome Biol.17, 122 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Kircher, M. et al. A general framework for estimating the relative pathogenicity of human genetic variants. Nat. Genet.46, 310–315 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Leeuw, C. A. de, Mooij, J. M., Heskes, T. & Posthuma, D. MAGMA: Generalized Gene-Set Analysis of GWAS Data. PLoS Comput. Biol.11, e1004219 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Liu, J. Z. et al. A versatile gene-based test for genome-wide association studies. Am. J. Hum. Genet.87, 139–145 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Chen, E. Y. et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinforma.14, 128 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information (929.6KB, pdf)
42003_2025_8778_MOESM2_ESM.pdf (88.2KB, pdf)

Description of Additional Supplementary Materials

Supplementary Data 1 (55.8KB, xlsx)
Supplementary Data 2 (18.2KB, xlsx)
Reporting summary (2.7MB, pdf)
Transparent Peer Review file (472.1KB, pdf)

Data Availability Statement

Individual level data are available from the Canadian Longitudinal Study on Aging. Interested researchers should submit an online application including a research proposal. Researchers who meet the access criteria will receive de-identified data via secure download. Full instructions are available at https://www.clsa-elcv.ca/data-access/. GWAS and XWAS summary statistics generated in this study are publicly available via Zenodo at 10.5281/zenodo.16897822. All other data supporting the findings of this study are available in Supplementary Data 1 and 2.

Scripts that were used in this study are available on https://github.com/Drogemoller-Lab/ARHL.

Articles from Communications Biology are provided here courtesy of Nature Publishing Group

RESOURCES