Article Text

Original article
Mitochondrial mutations associated with aminoglycoside ototoxicity and hearing loss susceptibility identified by meta-analysis
  1. Wu Jing1,
  2. Hao Zongjie1,
  3. Fu Denggang1,
  4. Hei Na1,
  5. Zhang Bin2,
  6. Zhou Aifen2,
  7. Hu Xijiang2,
  8. Yao Cong2,
  9. Dong Yunping3,
  10. Huijun Z Ring1,
  11. Brian Z Ring1
  1. 1Institute of Genomic and Personalized Medicine, College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, Hubei, China
  2. 2Wuhan Women and Children Hospital, Wuhan, Hubei, China
  3. 3Wuhan Bureau of Health, Wuhan, Hubei, China
  1. Correspondence to Professor Brian Z Ring, Institute of Genomic and Personalized Medicine, College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, Hubei 430079, China; bzring{at}


Background Genetic variations, including mitochondrial mutations, are important contributors to hearing loss, especially in children, and newborn genetic screens for hearing loss mutations are becoming increasingly common. Mitochondrial mutations have been linked with ototoxic responses to common antibiotics, therefore understanding the association of these mutations with hearing loss is of special importance. To address the usefulness of screening for these mutations in a clinical setting, we formed a collaboration of clinicians and geneticists to analyse the association of mitochondrial mutations with non-syndromic hearing loss, including the effect of ethnicity, audiological test methods and aminoglycoside exposure.

Methods This survey identified 122 variants in 43 studies that have been assessed for an association with hearing loss, and meta-analysis was performed on clinically relevant subsets. RNA folding and conservation analysis further explored possible relevance of these variants.

Results Among all studies, eight variants were found to have significant associations with hearing loss. A partially overlapping set of six variants had significant association with hearing loss when aminoglycoside exposure was assessed. Five of these variants predictive of sensitivity to aminoglycoside spatially co-localise in an RNA folding model. There was little effect of the audiological test method used to assess hearing loss on the association with the variants.

Conclusions Our results found a small set of studied variants had reproducible association with hearing loss, which will help clarify mutations useful in genetic screens for hearing loss. Several of the aminoglycoside exposure-associated mutations may co-localise on folded 12S rRNA, suggesting a functional association between these loci and aminoglycoside-induced hearing loss.

  • Genetic screening/counselling
  • Clinical genetics

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Hearing loss is a common disorder that affects approximately 10% of the global population, with 275 million people being moderately to severely affected, and 65 million suffering hearing loss during childhood.1 At birth 3% of newborns in developed countries and >6% in developing countries have detectable hearing problems.2 Most hearing loss is non-syndromic, but hearing loss can also be associated with other conditions. As early detection of hearing loss is critical in mitigating the profound effects that deafness has on language and social development,3 ,4 universal newborn hearing screening (UNHS) is increasingly common.5 Current screening is largely oriented towards identifying profound hearing loss; however, it may be desirable to expand the use of UNHS to identify children with mild hearing loss as even slight hearing impairment can effect language skills.6 An impediment to this is that current audiological examinations have significant failure rates, with both false positive and false negative diagnoses being prevalent.7 ,8 Though this can be substantially improved by repeated testing, identification of screening criteria could greatly increase the efficiency and applicability of UNHS and may be an important aid in identifying which individuals should receive more extensive early audiological testing.9 ,10 In addition, delayed-onset hearing loss is an important contributor to paediatric deafness. In a multiple independent study of large numbers of preschool children who had exhibited no hearing loss in newborn hearing screenings, approximately 1‰ were found to later exhibit permanent delayed-onset hearing loss.11 ,12 Testing for variants associated with hearing loss have been used to successfully identify children at risk for late-onset hearing loss despite having no detectable hearing loss as a newborn.13

In this regard, genetic screening of newborns for variants that may predispose towards non-syndromic hearing loss (NSHL) is being increasingly explored.14–17 Genetic and environmental factors both contribute to NSHL, with genetic variation contributing to at least half of all cases of profound congenital deafness.18 Several genes have been shown to play important roles in the manifestation of hearing loss, with GJB2, SLC26A4 and mitochondrial mutations accounting for most of the known variants. The maternally inherited mtDNA mutations associated with deafness are generally homoplasmic, meaning that the mtDNAs in a cell are identical. Mitochondrial mutations are of special interest in that some variants have been linked to sensitivity to ototoxic aminoglycoside antibiotics (for review, see ref. 19). This class of drugs includes neomycin, streptomycin, gentamycin and tobramycin, as other commonly used antibiotics. Treatment with aminoglycosides in some patients leads to bilateral permanent hearing loss. One study conducted in China, where aminoglycoside use is relatively widespread, found that nearly 30% of paediatric deaf subjects could be associated with exposure to aminoglycosides.20 However, mitochondrial mutations have been associated with hearing loss in the absence of the use of known ototoxic drugs, a complete explanation of how mitochondrial mutations lead to hearing loss is not available. It is possible that mitochondrial mutations interfere with metabolism and that insufficient energy production leads to degeneration of cochlear cells (for review, see ref. 21). However, the cochlea is not the most energy-dependent organ in the body, leading to the question of how penetrance is determined and why the ear is preferentially affected. Other environmental factors have been shown to play significant roles in hearing loss, primarily trauma and infections of the ear. In addition, mitochondrial haplotype and other non-mitochondrial genes have been shown to have modifying effects on the mitochondrial mutations.22

Though the importance of some mitochondrial mutations to hearing loss has been well established, with the 12S RNA (MT-RNA1) m.1555A>G mutation most well studied, many other variants have been identified.23 Therefore, we formed a collaboration of clinicians and geneticists to perform a meta-analysis of mitochondrial variants and their association with hearing loss in clinically relevant subgroups. The questions we wished to address were the prevalence of hearing loss-associated mitochondrial mutations in different ethnic groups, which polymorphisms on the mitochondrial DNA are associated with hearing loss and what is the pooled effect size of reported associations for the reported associations. When data were available, the interaction of exposure to aminoglycosides was also examined, as well as how the audiological method used to assess hearing loss affected the association of the identified mutations with hearing loss.

Materials and methods

Literature search strategy

Prior to initiation of the study, the analysis methodology was submitted to the PROSPERO register on December 2013 (registration ID: CRD42013006280). Articles indexed in PubMed or Web of Science before December 2013 were searched, without language restrictions, using the search terms: (polymorphism OR SNP OR CNV OR "copy number variation" OR mutation OR genetic OR InDel OR marker) AND ("hearing loss" OR deafness OR Deafness[MESH]) AND (cohort OR “case control” OR population) AND (mitochondria OR mitochondrial OR 12S OR 16S OR rRNA OR tRNA OR mtDNA OR mDNA OR NADH OR cytochrome). Five additional studies were added when the publications identified by the literature search referenced data included in other publications.

Inclusion and exclusion criteria

Case–control and cohort studies, including both population-based samples and other sampling methods, were included in this study. We included studies identifying hearing loss either through questionnaires or clinical examination. The review encompassed any defined and clearly identifiable genetic polymorphism found on mitochondrial DNA, including SNPs, insertion/deletions and CNVs. All participants in association studies were included, regardless of age. Studies that did not record possible exposure to aminoglycoside were included, as were studies that accepted only participants with known aminoglycoside exposure; however, these studies were appropriately flagged to allow proper analysis.

Participants from studies addressing comorbidities (eg, diabetes and deafness), from family studies or from studies with populations smaller than 10 were excluded from analysis. In addition, studies that did not provide complete case and control numbers for the assessed genotypes were not included.

Data extraction

Two reviewers independently screened publications for eligibility and disagreements resolved through discussion. Data were extracted from the agreed upon list of publications using a shared data form. Extracted data included study title, author names, year of publication, journal name, language of publication, location of study (country), ethnic composition, population selection methods, study size, numbers of cases and controls, polymorphism(s) tested, frequency of assessed genotypes in different groups, reported effect size, method of assessing hearing loss, exposure to aminoglycoside and reported variance in data.

Statistical analysis

The associations between all included mutations and hearing loss are presented as ORs with the corresponding 95% CIs and p values. Subgroup analysis was also performed by the method used to assess hearing loss and ethnicity, when this data were available. Fixed-effects models24 were used for any mutations with only one contributing study, and random-effects models (restricted maximum-likelihood estimator) were used for all other mutations. The significance of the pooled OR was determined using a Z test. To test whether the variability in the observed effect size is larger than expected based on sampling variability, heterogeneity was assessed by Cochran's Q test (Cochran, 1954) and I2 test.25 The heterogeneity was considered statistically significant if p<0.10. Possible study bias was tested by Egger’s regression and visualised via funnel plots. Meta-analysis and related procedures were performed using the metafor package in R. RNA folding was performed using Sfold, which uses as statistical sampling of RNA secondary structures from a Boltzmann ensemble of RNA secondary structures to predict folding.26 Sequence alignment employed ClustalW.27 All p values are presented as two-tailed, and values <0.05 were considered statistically significant.


Meta-analysis of overall data

From the 541 identified papers, 43 provided complete information allowing for further analysis (figure 1). These studies encompassed 122 mitochondrial variations on 11 genes, with over 36 000 total unique participants (table 1 and figure 2). Seven mutations were found to have significance in the meta-analysis combining all ethnicities, excluding studies where aminoglycoside exposure was an inclusion criterion. Of these mutations, seven occur in MT-RNR1 and one significant association was found for a mutation in mitochondrially encoded tRNA threonine (MT-TT) (table 2). All mutations showing significant association with hearing loss showed non-significant residual heterogeneity. Forest plots of variants included in table 2 are given in online supplementary figure S1, and variants with non-significant associations for which there were at least 10 contributing studies are shown in online supplementary figure S2. Egger's regression test showed no significant asymmetry of effect size compared with population (p=0.43), as did examination of a funnel plot (see online supplemental figure S3).

Figure 1

Search strategy and included study numbers.

Table 1

Variants included in this review

Figure 2

Map of mitochondrial mutations in the included studies.

Figure 3

Folding of 12S rRNA using an ensemble of RNA secondary structures to predict the folded structure. The entire RNA sequence was used to model folding, but only the regions near five of the six mutations shown to have an association with aminoglycoside related deafness are shown. (A) Region near MT-RNR1m.1555A>G and m.1494C>T. (B) Region near MT-RNR1m.839A>G, m.1095T>C and m.1107T>C.

Table 2

Variants showing significance for hearing loss

Subset analysis in ethnic groups

Due to the higher prevalence of NSHL, especially in paediatric populations, the majority of genotype association studies for mitochondrial mutations have been performed in Asia. Of the seven sites that showed significance in the overall population, two have not yet been assessed in Asia (MT-RNR1m.1557A>C and m.988G>A) and the MT-RNR1m.1494C>T just missed significance in an Asian population though was significant in the combined population (as discussed later, this mutation increases in significance when aminoglycoside exposure is considered). Only three of these mutations have been assessed in European populations ((MT-RNR1m.961ins/dels, m.1555A>G and m.988G>A), and all showed significance. In African populations, m.1555A>G did not show significance (OR=0.55, 95% CI −1.84 to 2.93); however, only two studies on African populations were included in this analysis.

MT-RNR1 961

Position 961 in MT-RNR1 can contain insertions of varying numbers of nucleotides, substitutions or deletions. The insertion and deletions mutations were grouped to allow for greater patient numbers (termed m.961ins/del in this study). No mutations at this position were significant in the combined populations, but the site did show a significant association with NSHL in European populations (OR 1.74, 95% CI 0.2 to 3.28). When assessed individually, no variant at position 961 is significant in any ethnic group subset analysis or when all combined patients are assessed. The most significant variant was MT-RNR1m.961delTinsC(2_7) in European populations (OR=1.69, 95% CI −0.49 to 3.87, p=0.12) and MT-RNR1m.961T>C in Asian populations, which had an opposite association with hearing loss (OR=–0.61, 95% CI −1.24 to 0.023, p=0.06).

Subset analysis on audiological tests

A variety of methods can be used to assess hearing performance. For this study, the methods were grouped as pure tone audiometry (PTA), CT, auditory brainstem response (ABR) and otoacoustic emission (OE). Most participants were assessed using multiple methods, and PTA, ABR and OE testing were the most common. The results were broadly consistent across hearing methodology (table 3). MT-RNR1m.1095T>C, which was not significant when using all data, is significant within PTA, ABR and OE assessed participants. In addition to subset analysis, mixed models including hearing tests as study-level covariates were used to determine effect size, and no additional variants were found to have a significant association with NSHL.

Table 3

Effect of hearing test on association

Aminoglycoside use

To address how aminoglycoside use affects an association between genotype and NSHL, individuals with known aminoglycoside exposure were assessed. When a Fisher's exact test is used to compare genotypes of deaf individuals with aminoglycoside exposure to those deaf individuals known to not have exposure, five MT-RNR1 variants are found to have a significant association (table 4). Of these five mutations, the m.1555A>G variant has the strongest association, as well as having the highest minor allele frequency. As most studies do not indicate whether participants were not exposed to aminoglycoside, this association was also assessed by performing the meta-analysis on deaf individuals with known aminoglycoside exposure compared with all normal controls. Five MT-RNR1 variants were found in this analysis, all with ORs in the same direction (table 5). m.1555A>G (OR=3.5, 95% CI 2.53 to 4.46) and C1494T (OR=2.47, 95% CI 1.04 to 3.91) had the strongest associations. One allele found with the Fisher's test (MT-RNR1m.1027A>G) was not significant in this analysis (p=0.10); this same allele barely achieved significance with the Fisher's test (p=0.048). An allele found via the meta-analysis but not via Fisher's test was MT-RNR1m.1107T>C (OR=1.4, 95% CI 0.21 to 2.58); insufficient data on aminoglycoside use were available to perform Fisher's test. These alleles had been primarily studied in Asian populations, and only m.1555A>G had sufficient data to assess via meta-analysis in a European population, where this allele was also significant (OR=3.14, 95% CI 0.81 to 5.48). Alleles that were present in the study but did not show significant results in the meta-analysis are presented in online supplementary table S1.

Table 4

Studies limited to cases receiving aminoglycoside

Table 5

Deafness in aminoglycoside users compared with all hearing study participants

To further examine the sites potentially important for aminoglycoside interaction, RNA folding modelling was performed. Of the six mutations identified as potential aminoglycoside interaction mediators, five variants localised to two regions (figure 3) m.1555A>G and m.1494C>T are potentially positioned opposite each other on a hairpin loop in the folded rRNA product, and mutation as either site may affect the folding of this region, as previously proposed.66 The mutations m.839A>G, m.1095T>C and m.1107T>C all lay within a small region of the folded RNA, which may be a critical site for aminoglycoside interaction. Alignment of the several 12S molecules shows a general high conservation of the gene (figure 4). The six sites involved with aminoglycoside interaction are more conserved than an average residue in the human 12S gene, but the conservation was not statistically significant.

Figure 4

Alignment of multiple 12S rRNA genes from diverse organisms. The regions surrounding the six mutations shown to have an association with aminoglycoside related deafness are shown and the variant site highlighted. A graphical representation of residue conservation is shown below each aligned region.


Hearing loss during early development can have profound effects on language and social development, but it is well established that these negative effects can be substantially abrogated if detected at an early date.3 ,4 For this reason, UNHS programmes are increasingly standard and have been shown to increase the frequency of newborns identified with hearing loss compared with physician or parental recognition and selective screening.5 As hearing tests have substantial failure rates in newborns7 ,8 and hearing loss is not always manifested until later in development, it is hoped that newborn genetic screening may be a useful tool in conjunction with UNHS to help identify at-risk individuals for more extensive audiological examination.

Mitochondrial mutations are known to play significant roles in susceptibility to NSHL. These mutations, in combination with other well-established non-mitochondrial mutations associated with high risk of hearing loss, may therefore be useful markers in identifying patients with hearing loss, especially in newborns. In addition, genetic screening for mitochondrial mutations may be useful in determining patients sensitive to aminoglycoside antibiotics. These antibiotics are commonly prescribed, especially in preterm neonates, because of their efficacy in combating sepsis.67 ,68 Because of their low costs and high efficacy, this class of antibiotics is more broadly prescribed in developing countries,69 thus efficient mechanisms of identifying patients where their use is contraindicated could be a great benefit. Therefore, our group of paediatricians and geneticists performed a review and meta-analysis of all available relevant and appropriate genetic studies. The goal was to determine which mitochondrial variants may be useful for newborn genetic screening, especially in countries where aminoglycoside use is widespread. It was also desired to determine whether there is information specifying how different audiological tests may interplay with a genetic screen, that is, would a certain gene be more or less applicable in a genetic screen depending on which audiological tests are employed in the newborn hearing screen.

In this survey, studies on 122 mutational sites on the mitochondrial DNA were identified in 11 genes. Most did not show a consistently significant association with hearing loss. Ten variants in the 12S ribosomal RNA gene (MT-RNR1) showed significance (m.961ins/del, m.1382A>C, m.1555A>G, m.1557A>C, m.839A>G, m.1009C>T, m.1494C>T, m.988G>A, m.1095T>C, m.1107T>C), as did one MT-TT variant (m.15927G>A). Five mutations were found to be significant when exposure to aminoglycoside was considered (m.1555A>G , m.839A>G, m.1494C>T, m.1095T>C, m.1107T>C), and one of these (m.1107T>C) was only significant when aminoglycoside exposure was considered.

Probably because of regional differences in the prevalence of paediatric NSHL, most of the alleles in this study have only been studied in Asian populations. The MT-RNR1 variants m.1382A>C, m.1009C>T and m.1494C>T, as well as the MT-TT variant m.15927G>A, were shown to be significant in Asian populations but have not yet been assessed in other regions. Similarly, the MT-RNR1 variant m.988G>A is significant in a European population but has yet to be assessed in an Asian population. Analysis of these variants in more populations may lead to a better understanding of the clinically useful variants for NSHL genetic screens.

The association of the identified mutations with NSHL was also measured in context with the audiological test used to assess hearing in the included studies. It is possible that hearing loss identified by different methods of assessing hearing loss may have different associations with the genetic variants. In this case, linking a genetic screen for hearing loss susceptibility with the most relevant method for the audiological newborn hearing screening programme would be important. Subset analysis and mixed models were used to determine effect size and significance of the variant's associations with hearing loss. However, though some differences were observed in the strength of association among the studied mutations between the different audiological testing methods, no striking inconsistencies were found (table 3). This analysis is somewhat confounded by multiple hearing tests being employed together, which was the norm in the included studies. PTA was the most commonly employed hearing test; very few studies did not employ PTA. ABR and OE, the other two most common tests, were less frequently used, almost always in conjunction with PTA. With this limitation in mind, we find no evidence to suggest that the method of audiological testing employed in the newborn screening would affect gene choice for a supplementary genetic screen for hearing loss susceptibility.

Aminoglycoside exposure is a significant cause of NSHL in paediatric patients in some regions, particularly in China.20 Mutations m.1555A>G and m.1494C>T have been associated in aminoglycoside-induced deafness in several studies, and these sites are near each other on the folded rRNA. Four other highly conserved MT-RNR1 mutations were seen in this analysis to also be associated in aminoglycoside-related NSHL, and three of these possibly spatially co-localise in the folded RNA. This might indicate a region that interacts directly with aminoglycoside or a region that affects how the other ribosomal proteins interact with aminoglycoside. The bases predicted to lie opposite 839 and 1095 (positions 846 and 855, respectively) have no reported human variants. These regions are also conserved across species (figure 4); however, there is a general high conservation of this gene sequence, so it is not possible to say that these regions are uniquely conserved. How all these sites might interact with aminoglycoside is a question to be further explored.

Though multiple variants are assessed in this analysis, no attempt at a correction to control for type I errors is made. When multiple studies were present for a single variant, random-effects models were employed; random-effects models are useful when attempting to estimate the effect size of a variant across different populations.70 As the findings presented here are likely to be useful to those interested in single mitochondrial mutations, correcting for other mutations included in this study would not be appropriate. However, in the consideration of a panel of multiple mutations for screening purposes based on this review, such a correction may be appropriate, and independent validation of the panel would be important.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • WJ and HZ are co-first authors.

    HZR and BZR are co-communication authors.

  • Contributors WJ, HZ, HZR and BZR participated in the design of the study. WJ, HZ and BZR performed the statistical analysis. FD and HN assisted with data preparation. ZB, ZA, HX, YC and DY provided clinical guidance, overview and contributed to data analysis. WJ, HZ, HZR and BZR wrote the manuscript. All authors read and approved the final manuscript.

  • Funding This study was partially funded by a grant from the Wuhan Bureau of Health.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement All relevant data are shared.