Relationship between CDH23 gene and risk of noise-induced hearing loss

Objective: To investigate the relationship between CDH23 gene and the risk of noise-induced hearing loss (NIHL). Methods: This was a case-control study. Noise-exposed workers worked in a steel factory in North China was recruited and been divided into two groups: the case group (BHFTA ≥ 40 dB) and the control group (BHFTA<25 dB). We analyzed the association among 18 single nucleotide polymorphisms (SNPs) in CDH23 and NIHL risk using the generalized multifactor dimensionality reduction (GMDR) method. Logistic regression was performed to analyze the main effects of SNPs and the interactions between CNE and SNPs adjusting cumulative noise exposure (CNE), smoking, drinking, physical exercise and hypertension. Results: In this study, 776 subjects of period I and 1117 subjects of period I+II were recruited. The results showed that subjects who carried the AA genotype of rs3802711possessed signicantly increased risk of NIHL than those carrying GG (OR: 2.71; 95% CI:1.15, 6.39) and GA+GG (OR: 2.54; 95% CI: 1.09, 6.00) in period I, respectively. For rs11592462, subjects carrying the GG genotype showed a signicantly increased risk of NIHL compared with the subjects. Signicant relationships were showed between rs10999947, rs3802711, rs10762480, rs3752751, rs3752752, rs3747867, and rs11592462 for NIHL overall and various CNE strata. There was no signicant association between the rs1227049 - rs3752752 rs10999947 - rs3752751 - rs10762480 - rs3802711 - rs11592462 - rs4747195 - rs4747194 - rs10466026 haplotypes and NIHL risk. Conclusions: The genetic variation in the CDH23 gene might play an important role in determining individual susceptibility to NIHL. In and linked to loss in waltzer 6 Previous studies revealed that genetic variations in CDH23 is a key determinant of age-related hearing loss and early-onset progressive hearing loss, mice carrying this susceptibility allele was more susceptible to noise. 6–10 Mutations in CDH23 in mice cause stereocilia disorganization and lead to deafness and vestibular disorders. In humans, mutations in CDH23 lead to both nonsyndromic and syndromic hearing loss. Existing evidence from epidemiological studies suggested an inconsistent association between genetic variations and risk of NIHL. 11 One study reported that the associations between 63 polymorphisms in CDH23 and NIHL were all negative in the Polish and Swedish population. 12 Other studies 13–14 described the positive association of CDH23 polymorphisms with NIHL. relationship between genetic variation in CDH23 genes and risk of NIHL in Chinese population we conducted study to investigate the relationship between CDH23 gene and the risk of (Applied Biosystems, Foster City, CA, USA). The whole analysis process was performed blind. The concordance rates for quality control samples were between 99% and 100% for all assays. All the SNPs in the controls were in Hardy-Weinberg equilibrium (HWE) (p >0.05) except that 2 SNPs (rs1227051 in period I and rs1227049 in period I + II) with minor allele frequency less than 10% were excluded from the nal analysis. And that 3 SNPs (rs1227051, rs10823829 and rs1227065) were not signicantly associated with NIHL risk in overall or different CNE stata and different noise exposure level were not veried and genotyped in period I + II. A total of 17 SNPs in CDH23 (rs3802720, rs7087735, rs1227049, rs10999947, rs3752752, rs3752751, rs10823829, rs1227065, rs10999978, rs3747867, rs17712523, rs10762480, rs3802711, rs11592462, rs10466026, rs4747194, rs4747195) of period I and 14 SNPs in CDH23 (rs3802720, rs7087735, rs10999947, rs3752752, rs3752751, rs10999978, rs3747867, rs17712523, rs10762480, rs3802711, rs11592462, rs10466026, rs4747194, rs4747195) of period I + II were included in the nal analysis.


Introduction
Noise-induced hearing loss (NIHL) is a complex disease resulting from the interaction between environmental and intrinsic factors, studies in environmental factors included noise, organic solvents, heat, vibrations, smoking and drinking. [1][2][3] However, when exposed to similar levels of noise and environmental factors, the morbidity of NIHL varies widely among workers, that is, some workers have had NIHL, others not. 4 Genetic variation could explain some of the inconsistent results.
CDH23 is one of the most interesting genes to be investigated in humans, which encoded cadherin 23, a component of the stereocilia tip links, which are thought to gate mechanotransduction channel in hair cells. 5 In animals, CDH23 is the rst and the only gene linked with predisposition to noise-induced hearing loss in waltzer mice. 6 Previous studies revealed that genetic variations in CDH23 is a key determinant of age-related hearing loss and early-onset progressive hearing loss, mice carrying this susceptibility allele was more susceptible to noise. [6][7][8][9][10] Mutations in CDH23 in mice cause stereocilia disorganization and lead to deafness and vestibular disorders. In humans, mutations in CDH23 lead to both nonsyndromic and syndromic hearing loss. Existing evidence from epidemiological studies suggested an inconsistent association between genetic variations and risk of NIHL. 11 One study reported that the associations between 63 polymorphisms in CDH23 and NIHL were all negative in the Polish and Swedish population. 12 Other studies [13][14] described the positive association of CDH23 polymorphisms with NIHL.
The relationship between genetic variation in CDH23 genes and the risk of NIHL in Chinese population remains unknown. Therefore, we conducted this study to investigate the relationship between CDH23 gene and the risk of NIHL.
1 Materials And Methods

Subjects
In this study, noise-exposed workers worked in a steel factory in North China was recruited and been divided into two groups: the case group (BHFTA ≥40 dB) and the control group (BHFTA<25 dB). The details had been described in previous trial 15 . This study was conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of our hospital. All participants had signed the informed consent.

Inclusion and exclusion criteria
Inclusion criteria: (1) participants who exposed to occupational noise higher than 80 dB and the cumulative time of noise exposure was ≥3 years; (2) age was older than 18 years old; (3) patients who have signed informed consent. Exclusion criteria: (1) subjects who served as air force or artillery, have had the history of head trauma, explosion deafness and familial deafness, or had a history of contagious diseases (mumps, measles and rubella) and treatment with an ototoxic drug (aminoglycoside); (2) subjects with tympanitis, Meniere's syndrome, conductive hearing loss, exaggerated hearing loss, feigned deafness, sudden deafness, toxic deafness, deafness caused by contagious diseases, tumors, autoimmunological diseases and others; (3) subjects whose pure-tone audiogram showed horizontal lines or near-horizontal lines, and subjects who had impairment of hearing loss in linguistic frequency that was more severe than that of hearing loss in high frequency.

Data collection
Participation was voluntary, and written informed consent was obtained from all participants. Those who signed consent were interviewed by trained study nurses at a convenient location using a standardized and structured questionnaire. Information on name, sex, date of birth, height, weight, noise exposure history (including air force and artillery), tenure, history of past diseases (head trauma, Measles, mumps and rubella, tympanitis, Meniere's syndrome, explosion deafness and familial deafness), smoking history, drinking history, medication history, physical exercise habits and other diseases which can induce the hearing loss were collected through in-person interview, and the data of hypertension and the results of pure-tone audiometry (PTA) were required through professional test. Otology morphological examination and otoscopy were requested, including bilateral auricle malformation, external auditory canal malformation and stenosis, tympanic membrane perforation, adhesion or calci cation, etc.

PTA detection
The audiometry was done using AS216 audiometer (Interacoustics AS Company, Danish). All the audiometric tests were performed among 6297 workers with standard procedures in quiet test rooms with a background noise level of <25 dB (A) by trained occupational health physician. The data of pure-tone air conduction hearing threshold tests were recorded at the frequencies of 500, 1000, 2000, 3000, 4000, and 6000 Hz after subjects stopped noise exposure for at least 48 h. The averages of 3000, 4000, and 6000 Hz were calculated as the threshold levels at high frequency for each ear. The hearing thresholds at speech frequency were calculated by the average of 500, 1000, and 2000 Hz for each ear. The , noise exposure levels were assessed from 8 a.m. to 4 p.m. during the subjects' working time at the representational sites of each type of work using Noise Dose Meters (NoisePro series, Quest Technologies, American). Noise exposure was evaluated with equivalent continuous dB(A)-weighted sound pressure levels (L Aeq,8h ). Besides, the previous recorded data on noise exposure levels of the factory were collected. Cumulative noise exposure (CNE) was calculated to determine the actual noise exposure for each subject based on every period of occupational history, which was de ned as: 16 where L Aeq,8hi is the equivalent continuous A-weighted noise exposure level in decibels normalized to an 8 h working day, occurring over the time interval Ti in years, with a total of n different noise level exposure periods (i.e., years spent working in different noise tasks/environments), and T ref : 1 year.

SNP Selection and Genotyping
A total of 18 SNPs in CDH23 genes were selected for genotyping based on the following criteria: minor allele frequencies more than 5%, laboratory evidence of function, or prior association with human disease studies. The detailed information of the 18 SNPs in period I and the 16 SNPs in Period I + II were presented in the Table 1.
Genomic DNA was extracted from blood samples using the DNA Extraction Kit from Biotech Firm. The SNP genotyping work was performed using SNPscan method, which has been described in detail elsewhere, 17 PCR products were sequenced by ABI3730XL DNA analyzer and results were analyzed by GeneMapper 4.1 software (Applied Biosystems, Foster City, CA, USA). The whole analysis process was performed blind. The concordance rates for quality control samples were between 99% and 100% for all assays. All the SNPs in the controls were in Hardy-Weinberg equilibrium (HWE) (p >0.05) except that 2 SNPs (rs1227051 in period I and rs1227049 in period I + II) with minor allele frequency less than 10% were excluded from the nal analysis. And that 3 SNPs (rs1227051, rs10823829 and rs1227065) were not signi cantly associated with NIHL risk in overall or different CNE stata and different noise exposure level were not veri ed and genotyped in period I + II. A total of 17 SNPs in CDH23 (rs3802720, rs7087735, rs1227049, rs10999947, rs3752752, rs3752751, rs10823829, rs1227065, rs10999978, rs3747867, rs17712523, rs10762480, rs3802711, rs11592462, rs10466026, rs4747194, rs4747195) of period I and 14 SNPs in CDH23 (rs3802720, rs7087735, rs10999947, rs3752752, rs3752751, rs10999978, rs3747867, rs17712523, rs10762480, rs3802711, rs11592462, rs10466026, rs4747194, rs4747195) of period I + II were included in the nal analysis.

Statistical Analysis
We used the software program SPSS 22.0 (IBM, Chicago, USA) to conduct the statistical analysis. Before analysis, Hardy-Weinberg equilibrium (HWE) test was checked for each SNP among control subjects using χ 2 -test. Continuous variables were expressed as mean ± standard deviation (SD) while categorical variables were expressed as frequencies (%). Paired Samples t-test was used for two comparisons when each datum conformed to normal distribution, while the non-normally distributed continuous data were compared using non-parametric tests. The counting data were tested by chi-square test. Four genetic models were used, that were additive-inheritance model, dominant-inheritance model, recessive-inheritance model and co-dominant-inheritance model. Unconditional logistic regression model was used to estimate the odds ratios (ORs) and 95% con dence intervals (CIs) for associations between SNPs, risk of NHL, different CNE strata (CNE<97dBA year and CNE≥97dBA year) and different noise level (noise level<85dB and noise level≥85dB) in different genotype strata after adjusting for the potential confounding factors such as smoke, drink, hypertension, physical exercise and CNE. And Bonferroni correction was performed to control for multiple testing. Haploview was used to investigate the linkage disequilibrium (LD) between the SNPs (Haploview version 4.2 software, http://www.broad.mit.edu/mpg/haploview). Generalized multifactor dimensionality reduction (GMDR) software (version 0.9, http://sourceforge.net/projects/gmdr/) was applied for further detecting best multilocus model associated with NIHL after adjustment for covariates including smoke, drink, blood pressure, physical exercise and CNE. Finally, 1000 permutations were performed to get a permutated value of these models.

The general Characteristics
In this study, 776 subjects of period I and 1117 subjects of period I+II were in included. The concrete characteristic of the subjects were presented in the Table 2. The results showed that no statistically signi cant differences were seen between case and control individuals in the distribution of age, tenure exposure to noise, CNE, drinking status, blood pressure in both period I and period I + II (p> 0.05), and so physical exercise in period I (p> 0.05), but for period I + II, the results of physical exercise was just the opposite (p 0.05). However, the BHFTA in the case group was signi cantly greater than that of the control subjects both period I (51.9±9.1 vs. 19.2±7.5, p<0.05) and Period I + II (51.3±9.0 vs. 18.6±8.2, p< 0.05). Besides, the risk of NIHL in smokers both above two period group increased signi cantly than that of non-smokers (p< 0.05). The variables (CNE, smoking, drinking, physical exercise and blood pressure) were further adjusted in the unconditional logistic regression.

Associations of CDH23 Variants and the Risk of NIHL
As shown in Table 3 (presented only for the SNPs with signi cant results, the whole results of associations of CDH23 variants and the risk of NIHL were detailed in supplementary Table S1), after adjusting for CNE, smoking, drinking, physical exercise and blood pressure, subjects carrying the AA genotype of rs3802711were signi cantly associated with a signi cantly increased risk of NIHL than those carrying GG (adjusted OR: 2.71; 95% CI:1.15, 6.39) and GA+GG (adjusted OR: 2.54; 95% CI: 1.09, 6.00) in period I, respectively. For rs11592462, subjects carrying the GG genotype showed a signi cantly increased risk of NIHL compared with the subjects carrying CC (adjusted OR: 2.40; 95% CI: 1.26, 4.56) and CG+CC (adjusted OR: 2.47; 95% CI:1.31, 4.66) both in period I and period I + II. However, no signi cant differences were detected in genotype of the other 15 SNPs of CDH23 in period I and the other 13 SNPs in period I + II between case and control subjects (p>0.05). However, none of the associations remained statistically signi cant after Bonferroni adjustment.

Strati cation Analysis of CDH23 by Noise Exposure Level and CNE
The results of associations between CDH23 SNPs and NIHL risk strati ed by noise exposure level and CNE were listed in the However, none of the associations remained statistically signi cant after Bonferroni adjustment.

Association of CDH23 Haplotypes with NIHL Risk
Haplotypes were inferred based on observed genotypes by using Haploview software, for period I and period I+II, rs1227049, rs3752752, rs10999947, rs3752751, rs10762480, rs3802711, rs11592462, rs4747195, rs4747194 and rs10466026 ten SNPs constructed the haplotypes in period I, but the results for period I suggested that there was no signi cant association between the rs1227049-rs3752752-rs10999947-rs3752751-rs10762480-rs3802711-rs11592462-rs4747195-rs4747194-rs10466026 haplotypes and NIHL risk, but the results for period I+II was the opposite, after adjusting for CNE, smoking, drinking, physical exercise and blood pressure, the detailed signi cant association suggested that the decreased NIHL risk was observed for the haplotype of GGCCCGCAAT (adjusted OR: 0.66; 95% CI: 0.43, 0.99) (See Table S2). After applying the Bonferroni correction, all the associations were no longer statistically signi cant.

Evaluation of the Interaction Effect between CDH23 SNPs
GMDR was performed to reveal the interactions between SNPs. None of signi cant SNP-SNP interactions were found in Period I+II of this study (P>0.05, see details in Table S3).

Discussion
Our study was the rst comprehensive analysis between genetic polymorphisms in CDH23 genes and risk of NIHL. In this study, SNPs located in the CDH23 genes in period I (17 SNPs) and period I + II (14 SNPs), the SNPs in period I with the signi cant and borderline signi cant increased or decreased NIHL risk were chosen to verify the association between gene variation in CDH23 and NIHL risk in period I + II. Signi cant associations were observed for rs10999947, rs3802711, rs11592462, rs10762480, rs3752751, rs3752752, rs3747867 for NIHL overall and/or various CNE strata and noise exposure level.
For overall, not only period I and but also period I + II, the study suggested that CDH23 polymorphism (rs11592462) may have increased NIHL risk which was consistent with the Polish population study, 18 and CDH23 polymorphism (rs3802711) may have increased NIHL risk in period I which was in accordance with the previous study. 19 For strati ed analysis by noise exposure level and/or CNE, when noise exposure levels ≥ 85dB(A) or when CNE ≥ 97dB(A)·year, the study suggested that CDH23 polymorphism (rs11592462) may have increased NIHL risk in period I + II, so did the CDH23 polymorphism (rs3802711) and the CDH23 polymorphism (rs10762480) in period I and period I + II.
In conclusion, the study suggested that CDH23 polymorphisms (rs11592462, rs3802711 and rs10762480) were signi cantly associated with NIHL risk both period I and period I + II, which was consistent with previous study, [19][20] which suggested that CDH23 gene polymorphisms play an important role in the development of NIHL, which was proved by the analysis of the association between the CDH23 haplotypes and NIHL risk.
CDH23 plays a necessary long-term role in the normal marshalling course of the ciliary bundle of the cochlea, 20-23 CDH23 mutations were rstly found to be associated with susceptibility to noise induced hearing loss in the population and the study recommended that CDH23 could be an early indicator of hearing loss in routine screening. 11 Although some studies did not manifest the association between gene variations in CDH23 and NIHL risk, 12,24 however recent one study suggested that CDH23 gene variations could increase NIHL risk, 14 which is consistent with the present study results.
The study has several strengths. Firstly, the study is a NIHL cohort population-based case-control study, in which it has a comprehensive data of noise exposure than previous studies. Secondly, the "cases" of this study were chose by the diagnosis of occupational noise-induced deafness (GBZ 49-2014) and the "controls" of this study were diagnosed by the same standard with the "cases", so the results of this study were different from the previous studies 11,14 in which study population was chose whether susceptible to noise or not and, to some degree in hearing loss, the results of this study were more valuable and objective than those studies, in that subjects of susceptibility of those studies might have normal hearing loss. Thirdly, the power size of the study in "case" of hearing loss was bigger than previous study. Lastly, the study divided two periods and the results of period I were veri ed in period II, and then analyze the results of period I + II with a big power size, thus the results appears convincing.
This study also has some limitations that should be considered. Firstly, when the Bonferroni correction was applied in our study, the signi cance of the ndings was overwhelming; However, it is recognized that replication of ndings in independent populations is much more important than obtaining highly signi cant p-values. Secondly, the workers might be exposed to noise in other places, such as the community, but this was too complicated to take into consideration; but one recent review 25 based on different high noise exposure groups suggested that for high intensity noise exposure study, the noise intensity of daily life has little effect on the results of the study. Lastly, this study focused on a speci c pathway, additional SNPs should be tested in the future.

Conclusions
The genetic variation in the CDH23 gene might play an important role in determining individual susceptibility to NIHL. The results provide new insight into the pathogenesis and early prevention of noise-induced hearing loss.

Declarations
Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of Henan Medical College.

Consent to publication:
All authors nal approval of the version to be published, and informed consent was obtained from all participants.

Competing interests:
All of the authors had no any personal, nancial, commercial, or academic con icts of interest separately.