Associations of genetic variation in E3 SUMO-protein ligase CBX4 with noise-induced hearing loss

Abstract

Noise-induced hearing loss (NIHL) is a multifactorial disease caused by environmental, genetic and epigenetic variables. SUMOylation is a post-translational modification that regulates biological processes. The objective of this study was to determine the link between genetic variation in the chromobox 4 (CBX4) and the risk of NIHL. This study applied a case–control design with 588 cases and 582 controls, and the sample was predominantly male (93.76%). The T allele of CBX4 rs1285250 was found to be significantly linked with NIHL (P = 0.002) and showed strong associations in both the codominant and recessive models (TT versus CC, P = 0.005; TT/TC versus CC, P = 0.009). By constructing a mouse model of hearing loss because of noise exposure, changes in hearing thresholds were observed in noise-exposed mice, along with a decrease in the number of cochlear hair cells. Furthermore, noise promotes cochlear hair cell apoptosis by inducing SP1/CBX4 pathway activation. Further functional studies demonstrated that SP1 has an influence on the promoter activity of the CBX4 rs1285250 intron, with the promoter activity of the T allele being higher than that of the C allele. Knockdown of transcription factor SP1 reduced the expression of CBX4 expression and simultaneously reduced apoptosis in HEI-OC1 cells. Together, our findings have shown that CBX4 genetic polymorphism rs1285250 T-allele was associated with increased risk of NIHL and might be used as biomarkers for male workers exposed to noise. Furthermore, we speculate that the CBX4 of rs1285250 T-allele leads to a stronger potential enhancer activity from a predicted gain of stronger SP1 binding.

Introduction

Occupational noise is the second most important risk factor in the workplace after occupational injuries and is common in today’s world (1,2). Worldwide, 16% of the disabling hearing loss in adults (over 4 million disability-adjusted life year) is attributed to occupational noise, ranging from 7 to 21% in the various subregions (3,4). Long-term noise exposure, in particular, can cause irreparable hearing loss, although it can be efficiently prevented (5). Even when subjects were exposed to similar noise levels, they had varying degrees of noise-induced hearing loss (NIHL) risk (6). These findings suggested that genetic susceptibility, as well as its interplay with environmental variables, may play a role in the development of NIHL. Previous genetic investigations in humans have shown that polymorphisms in HOGG1 (7), NFE2L2 (6), HSP70 (8) and CDH23 (9) are linked to the risk of NIHL.

Post-translational modifications (PTMs) are important molecular processes that change proteins after they have been synthesized and modulate their eventual functional characteristics by modifying their interaction potential, stability, activity or localization (10,11). Protein PTM changes have been associated with the pathogenesis of a variety of human illnesses, including deafness (12). Furthermore, excessive oxidative stress in the cochlea is associated with the pathogenesis of NIHL, implying that reducing oxidative stress is an effective method for reducing the prevalence and development of NIHL. A key concern in protecting the structure and function of the cochlea is how to enhance cochlear resistance to oxidative stress (13,14).

SUMOylation plays an important role as a key PTM event in the regulation of oxidative stress and the development of hearing loss (12). SUMOylation can potently reduce reactive oxygen species (ROS) production from NOX1 to NOX5 (15), whereas SUMOylation of PRDX6 promotes its degradation and consequently attenuates ROS scavenging (16). Furthermore, oxidative stress influences SUMOylation through a variety of direct and indirect pathways. By forming a disulfide link between the catalytic cysteines of the SUMO E1 and SUMO E2 enzymes, H₂O₂ can reversibly inhibit SUMOylation (17). The activities of the SUMO E3 ligase can be influenced by ROS. H₂O₂ treatment, for example, improves the interaction of the E3 ligase PIAS4 with its target NEMO (18). Furthermore, SUMOylation can also affect the proper formation of the otic vesicle by regulating the expression of PAX8 (12). Chromobox 4 (CBX4) is the only protein in the family that possesses enzymatic activity and can act as a SUMO E3 ligase for SUMO modification (19). The increased expression of CBX4 caused by oxidative stress resulted in a significant downregulation of cell viability and upregulation of CASPASE-3 activity, suggesting that SUMOylation inhibits proliferation and promotes apoptosis and that de-SUMOylation reverses this process (20).

Despite previous evidence connecting SUMOylation to the oxidative stress response, research on their impact on hearing loss is limited. We conducted cross-sectional research in a Chinese population to analyze the association between SUMO tagSNPs and CBX4 rs1285250 T allele risk and determine the single nucleotide polymorphisms, SNPs’ biological effect.

Results

Study populations

A total of 1170 individuals were included in the study, including 588 in the case group and 582 in the control group, and the sample was predominantly male (93.76%; Supplementary Material, Table S3). The sex ratios of male and female subjects in the case and control groups were very similar [549 (93.37%) versus 548 (94.16%) for male and 39 (6.63%) versus 34 (5.84%) for female]. No significant differences in age, gender, noise exposure level, duration of noise exposure, cigarette smoking and alcohol consumption were found between the two groups. However, the average hearing threshold was considerably higher in the case group than in the control group (P < 0.001).

Selected SNPs

According to the screening criteria, 31 SNPs of SUMO1 activating enzyme subunit 1 (SAE1), ubiquitin like modifier activating enzyme 2 (UBA2), RAN binding protein 2 (RANBP2), CBX4 and ubiquitin conjugating enzyme E2 I (UBE2I) genes were included in this study. Table 1 shows the basic information about all SNPs, including gene, location, alleles and Hardy–Weinberg equilibrium (HWE) values. Genotyping of all SNPs was performed using multiplex PCR and second-generation sequencing techniques, and among these 31 SNPs, only 2 SNPs of CBX4 (rs1285249 and rs1285250) showed statistically significant differences in genotype distribution between the case and control groups (P < 0.05), so the subsequent association analysis was focused on these 2 SNPs to find out possible mechanisms involved in the phenotypic changes in CBX4.

Table 1

Selected SNPs and Hardy–Weinberg test

Gene	SNP	Chromosome	Functional consequence (intron number, total, ranked)	MAF		P for HWEb	Allele	NIHL	Control	P valuec	P valued
				Control	Databasea		A1/A2	A1A1/A1A2/A2A2
SAE1	rs8107443	19:47204170	Intron variant (6, 6)	0.241	0.228	0.970	G/T	42/209/312	30/209/319	0.368	0.425
	rs10424953	19:47202026	Intron variant (6, 6)	0.250	0.228	1.000	A/G	44/220/316	32/216/312	0.448	0.523
	rs309184	19:47133570	Intron variant (6, 1)	0.417	0.451	0.524	G/T	99/259/170	87/288/179	0.383	0.393
	rs8102445	19:47134351	Intron variant (6, 1)	0.315	0.311	0.353	G/T	278/248/39	255/241/52	0.260	0.316
	rs177918	19:47131056	Synonymous variant	0.116	0.126	0.080	G/T	420/136/19	433/108/10	0.058	0.062
	rs536759698	19:47132890–5	Intron variant (6, 1)	–	0.252	–	–	–	–	–	–
CBX4	rs112193822	17:79840734	Regulatory region variant	–	0.476	–	–	–	–	–	–
	rs1285243	17:79840307	Regulatory region variant	0.210	0.223	0.511	C/T	12/137/257	12/100/183	0.718	0.764
	rs1285249	17:79836344	Intron variant (4, 3)	0.215	0.180	0.537	C/G	9/171/395	26/190/348	0.002	0.005
	rs1285250	17:79836199	Intron variant (4, 3)	0.215	0.180	0.428	C//T	9/171/399	25/193/348	0.002	0.004
	rs200337966	17:79834771	Missense variant	–	0.141	–	–	–	–	–	–
	rs202078528	17:79834773	Missense variant	–	0.141	–	–	–	–	–	–
	rs4889898	17:79824605	Intron variant	0.307	0.296	0.9331	C/A	257/205/54	266/247/49	0.310	0.441
RanBP2	rs826556	2:108784803	3 Prime UTR variant	–	0.029	–	–	–	–	–	–
	rs700875	2:108772668	Intron variant (28, 22)	0.040	0.029	0.627	A/T	535/43/1	525/41/2	0.831	0.823
	rs1614285	2:108777559	Intron variant (28, 25)	0.194	0.029	0.619	A/T	1/44/540	2/40/528	0.793	0.801
	rs1649329	2:108776285	Intron variant (28, 24)	0.038	0.029	0.596	A/G	540/44/1	527/39/2	0.763	0.777
	rs1478517	2:108894481	3 Prime UTR variant	0.077	0.068	0.868	A/G	486/66/4	488/82/3	0.445	0.470
	rs2308203	2:108784836-9	3 prime UTR variant	–	0.049	–
UBE2I	rs7204003	16:1327362	Intergenic variant	0.252	0.214	0.652	A/G	44/222/320	38/202/311	0.805	0.763
	rs12925270	16:1325936	3 Prime UTR variant	0.110	0.102	0.384	A/C	9/115/461	10/103/444	0.841	0.846
	rs2281226	16:1314766	Intron variant (6, 3)	0.094	0.112	0.552	A/C	1/84/486	4/91/433	0.177	0.198
	rs11248866	16:1315340	Intron variant (6, 3)	0.062	0.112	0.581	A/G	493/87/2	387/53/1	0.375	0.398
	rs4984809	16:1316047	Intron variant (6, 4)	–	0.112	–	–	–	–	–	–
	rs761059	16:1324523	Intron variant (6, 6)	0.184	0.180	0.875	A/G	17/173/367	19/173/382	0.925	0.928
	rs4610	16:1314364	Synonymous variant	–	0.049	–	–	–	–	–	–
UBA2	rs7259160	19:34469074	Synonymous variant	0.198	0.150	0.995	A/G	18/179/372	22/175/357	0.758	0.760
	rs7258977	19:34468907	Intron variant (16,16)	0.227	0.184	0.828	A/G	29/193/344	31/190/334	0.937	0.940
	rs1559196	19:34460314	Intron variant (16,13)	0.226	0.184	0.861	C/T	29/196/353	31/194/342	0.930	0.934
	rs7252173	19:34455159	Intron variant (16,12)	0.221	0.184	0.953	A/G	28/188/339	26/178/316	0.992	0.999
	rs11673331	19:34461598	Intron variant (16,14)	0.226	0.184	0.930	C/T	351/200/29	342/195/31	0.941	0.937

^aMAF from the HapMap database (http://www.hapmap.org).

^bP-value for the Hardy–Weinberg test.

^cχ² test, P-value (two-tailed).

^dAdjusted for age, sex, smoking and drinking.

Association analysis of CBX4 gene polymorphisms with NIHL susceptibility

Table 2 illustrates the genotypes and allele distributions of CBX4 in 1170 participants. After adjusting for age, gender, cigarette smoking and alcohol consumption, logistic regression analysis revealed that the risk of NIHL in GG/GC allele carriers was 2.94 times higher than in CC allele carriers in the rs1285249 recessive model [95% CI (1.36–6.37), P = 0.006]. In the codominant model of rs1285250, the risk of NIHL in TT allele carriers was 3.09 times higher than in CC allele carriers [95% CI (1.41–6.74), P = 0.005]. In the allele model, those with the T allele had a 1.39 times higher risk of NIHL than those with the C allele [95% CI (1.13–1.72), P = 0.002].

Table 2

Distribution of four SNPs and associations with NIHL

Genetic models	Genotypes	Cases (n = 588)	Controls (n = 582)	Pa	Pb	Adjusted OR (95% CI)b	Holm	SidakSS	SidakSD
rs1285249		n = 575 (%)	n = 564 (%)
Codominant	GG	395 (68.7)	348 (61.7)	0.002		1.00 (Ref.)
	GC	171 (29.7)	190 (33.7)		0.070	1.26 (0.98–1.63)
	CC	9 (0.16)	26 (0.46)		0.003	3.18 (1.46–6.91)
Dominant	GG	395 (68.7)	348 (61.7)	0.013		1.00 (Ref.)
	GC/CC	180 (31.3)	216 (38.3)		0.015	1.36 (1.06–1.74)
Recessive	GG/GC	566 (98.4)	538 (95.4)	0.003		1.00 (Ref.)
	CC	9 (0.16)	26 (0.46)		0.006	2.94 (1.36–6.37)
Alleles	G	961 (83.6)	886 (78.5)	0.002		1.00 (Ref.)
	C	189 (16.4)	242 (21.5)		0.003	1.38 (1.12–1.70)	0.003	0.005	0.003
rs1285250		n = 579 (%)	n = 565 (%)
Codominant	TT	399 (68.9)	348 (61.5)	0.002		1.00 (Ref.)
	TC	171 (29.5)	193 (34.1)		0.044	1.30 (1.01–1.67)
	CC	9 (0.16)	25 (0.44)		0.005	3.09 (1.41–6.74)
Dominant	TT	399 (68.9)	348 (61.5)	0.008		1.00 (Ref.)
	TC/CC	180 (31.1)	218 (38.5)		0.009	1.38 (1.08–1.77)
Recessive	TT/TC	570 (98.4)	541 (95.6)	0.005		1.00 (Ref.)
	CC	9 (0.16)	25 (0.44)		0.009	2.84 (1.30–6.17)
Alleles	T	969 (83.7)	889 (78.5)	0.002		1.00 (Ref.)
	C	189 (16.3)	243 (21.5)		0.002	1.39 (1.13–1.72)	0.003	0.003	0.003

^aχ² test, P-value (two-tailed).

^bAdjusted for age, sex, smoking and drinking in logistic regression model. Drinking: based on past year alcohol habits, whether wine, liquor or beer and excluding occasional drinking on holidays. The alcohol consumption status was categorized into three groups: non-drinkers (those who did not consume alcohol even once a month), occasional drinkers (those who consumed at least one glass of alcohol every month) and frequent drinkers (those who consumed more than once per week).

Interaction between gene polymorphisms and environmental factors of noise exposure

In a dominant model, the effects of the rs1285249 and rs1285250 genotypes on a number of NIHL risk factors were investigated (Supplementary Material, Table S4). Participants with the GG genotype had a higher risk of NIHL than those with the GC/CC genotype after adjusting for confounding variables such as age, sex, smoking and alcohol consumption by a binary logistic regression model for noise-exposed workers working ≤ 15 years (P = 0.039). The genotype distributions of rs1285249 were substantially different between the case and control groups for noise exposure levels > 85 dB (A) (P = 0.004). Similarly, it can be concluded that there was an interaction between rs1285250 and the duration of noise exposure and intensity (P < 0.05).

Haplotype analysis

Table 3 shows the findings of the haplotype analysis. From 99% of the haplotype variations obtained from four SNPs, four major haplotypes (frequency > 1%) were selected (Supplementary Material, Fig. S1). Haplotype ATGT (rs4889898–rs1285243–rs1285249–rs1285250; OR = 1.29, CI 95% = 1.01–1.65, P < 0.05) was a risk factor for NIHL.

Table 3

Frequencies of inferred haplotypes in cases and controls and their associations with NIHL risk

Haplotypesa	Case (%)	Control (%)	Chi2	Pb	OR (95% CI)c	Pc
CTGT	264 (22.45)	225 (19.33)	3.44	0.064	1.22 (0.999–1.50)	0.051
ATGT	168 (14.29)	132 (11.34)	4.54	0.033	1.29 (1.01–1.65)	0.039
CCGC	92 (7.82)	84 (7.22)	0.31	0.579	1.08 (0.80–1.47)	0.615

^aAlleles of haplotypes were arrayed as rs4889898–rs1285243–rs1285249–rs1285250.

^bTwo-sided χ2 test.

^cAdjusted for age, sex, smoking and drinking in logistic regression model. Drinking: based on past year alcohol habits, whether wine, liquor or beer and excluding occasional drinking on holidays. The alcohol consumption status was categorized into three groups: non-drinkers (those who did not consume alcohol even once a month), occasional drinkers (those who consumed at least one glass of alcohol every month) and frequent drinkers (those who consumed more than once per week).

Gene–gene interaction analysis

In this study, the multifactor dimensionality method was used to analyze the interactions between genes. The four SNPs of the CBX4 gene were included in GMDR v0.9 software for analysis while corrected for age, gender, cigarette smoking and alcohol consumption. The results showed that in Table 4 and Supplementary Material, Figure S2, all four models were statistically significant (P < 0.05), indicating that rs4889898, rs1285243, rs1285249 and rs1285250 model was related to an increased risk of NIHL.

Table 4

MDR analysis of the interactions among the four SNPs

Model	Training Bal Acc	Testing Bal Acc	Sign test (P)	CV consistency
rs1285243	0.5924	0.5876	10 (0.0010)	10/10
rs1285243–rs1285249	0.5984	0.5856	10 (0.0010)	8/10
rs4889898–rs1285243–rs1285249	0.6088	0.5963	10 (0.0010)	9/10
rs4889898–rs1285243–rs1285249–rs1285250	0.6096	0.5954	10 (0.0010)	10/10

Comparison of SP1 binding luciferase activity of T and C allele of the rs9395890

To determine whether specific biological pathways or functions play a role in NIHL development, we used bioinformatics software to investigate the potential for SNPs in the identified hearing loss risk loci to affect the binding of transcription factors. We investigated two rs1285249 and rs1285250 variants that are thought to affect transcription factor binding. The SNP rs1285249 was anticipated to affect differential binding by HEY1, MTF1 and NRSF using multiple transcription factor motif analyses. Furthermore, the SNP rs1285250 altered the putative binding sites of the transcription factors of SP1 and had the highest score for binding to SP1 (Supplementary Material, Fig. S3). We next investigated whether SP1 binds to the putative SP1 binding site of CBX4 (rs1285250). The rs1285250 allele T increased the luciferase activity, whereas the allele C reduced the luciferase activity (P < 0.001). Based on these findings, the rs1285250 appears to be located in the center of a transcription factor SP1 consensus binding site consensus. When the C allele is present, we speculate that SP1 binding to this location may be weaker than when the T allele is present (Fig. 1).

Figure 1

Luciferase reporter assay. SP1-OE means SP1 overexpression. CBX4 WT means rs1285250 T allele, whereas the MT means rs1285250 C allele. Relative luciferase activity was expressed by the ratio of Firefly/Renilla activity. Data were presented as the mean ± SD [average of 3 replicates (± STD)]. (asterisk) Compared with SP1-NC +CBX4 WT. (hash) Compared with SP1-OE +CBX4 MT. ^*P < 0.05 and ^**P < 0.01. #P < 0.05 and ##P < 0.01.

Auditory brainstem response test and immunofluorescence

Hearing function in C57BL/6 mice was examined using auditory brainstem response (ABR) tests. The noise-exposed group’s average thresholds (at 4, 8, 16, 24 and 32 kHz) were considerably higher than the control group’s (Fig. 2a). A single row of inner hair cells (IHCs), three rows of outer hair cells (OHCs) and a number of interstitial support cells make up the organ of Corti, which sits atop the basilar membrane. After the cochlear basement membrane was immunofluorescence stained, the hair cell damage degree was shown in Figure 2b. It can be clearly seen that the three rows of OHCs and one row of IHCs in the cochlea of the control mice were neatly arranged with no significant loss, whereas the hair cells in the top segment of the cochlear basement membrane of the noise-exposed mice were disordered, and different degrees of hair cell loss appeared in the apex, middle and bottom segments, with the middle part showing the most serious hair cell loss. These data indicate that the model of NIHL was successfully constructed in C57BL/6 mice.

Figure 2

ABR measurement and cell loss patterns in cochleae of the mice. (a) ABR threshold between control and noise exposure group (n = 6 in each group). Average of 3 replicates (± STD), ^**P < 0.01 compared to the controls. (b) Hair cell loss (white arrows) was found in noise exposure group, Scale bar: 20 μm. (c) Quantification of HC in the apical, middle, and basal turns of the cochlea in mice (n = 6 in each group). Error bars indicate as means ± SD. ^*, P < 0.05; ^**, P < 0.01. Scale bar, 10 μm. ABR, auditory brainstem response; HC, hair cell; SD, standard deviation.

Effects of noise on GSH-Px, MDA and SOD

To further explore the pathways underlying NIHL, the activities of antioxidant enzymes were measured. Supplementary Material, Figure S4 presents the noise-exposed group that showed a significantly decreased GSH-Px activity (P < 0.05) and increased MDA level (P < 0.05) compared with the control group. Between the two groups, there were no significant variations in SOD activity.

Noise induces SP1/CBX4 pathway activation in cochlear tissues

RT-qPCR was used to determine the expression of Sp1 and Cbx4 mRNA, whereas the Elisa test kit was used to determine the protein expression of SP1 and CBX4. Figure 3a demonstrates that the noise-exposed group had higher levels of Sp1 and Cbx4 mRNA expression. The protein expression of SP1 and CBX4 was also significantly higher in the noise-exposed group (Fig. 3b), and the CASPASE-3, Cleaved-CASPASE 3 and BAX were significantly increased in the cochlear tissues of exposed mice (P < 0.01), whereas the content of anti-apoptotic protein BCL-2 was significantly reduced (P < 0.01, Fig. 3c and d). This suggests that noise enhances Cbx4 expression by inducing SP1 expression and thus promotes apoptosis.

Figure 3

(a) Quantitative real-time RT-PCR analysis of sp1 and cbx4 gene expression in the C57BL/6 mouse cochlea following noise exposure. (b, c) SP1, CBX4, apoptotic protein expression levels were measured by ELISA. Data presented as mean ± SD, average of 3 replicates (± STD). ^*P < 0.05 and ^**P < 0.01 relative to the control group.

Figure 4

(a, b) The expression of sp1 and cbx4 genes after treatment with siRNA targeting sp1. Values are expressed relative to those for the internal control, β-actin. (c–e) The expression level of SP1, CBX4, apoptotic protein evaluated by western blot. (f–g) Flow cytometric analysis was employed to detect the effect of SP1 knockdown on HEI-OC1 cell apoptosis. Error bars represent the mean ± SD of at least three independent experiments. ^*P < 0.05, ^**P < 0.01 versus control group.

Transcription factor SP1 promotes cell apoptosis by enhancing Cbx4 expression in HEI-OC1

We then investigated the mechanism of SP1-mediated Cbx4 activation. SP1 was knockdown by siRNA in HEI-OC1 cells. Low SP1 expression cell line demonstrated a significant decrease in SP1 expression (P < 0.01, Fig. 4a). In addition, the mRNA and protein expression levels of Cbx4 were lowered in a low SP1 expression cell line (P < 0.01, Fig. 4b and c). These findings suggest that transcription factor SP1 positively regulated CBX4 expression. The flow cytometry findings for apoptosis demonstrated that downregulation of SP1 dramatically decreased apoptosis in comparison with the si-NC group (Fig. 4f–g). The expression of CASPASE-3, Cleaved-CASPASE 3 and BAX were significantly decreased (P < 0.01), whereas the expression of anti-apoptotic protein BCL-2 was significantly increased (P < 0.01; Fig. 4c–e) in HEI-OC1 cells transfected with si-Sp1, indicating that low CBX4 expression inhibits HEI-OC1 apoptosis.

Discussion

The current findings suggest an association between the SUMO E3 ligase (CBX4) and NIHL. The rs1285250 T allele is associated with an increased risk of NIHL in male workers, whereas the SP1 transcription factor exhibits a different risk of NIHL by binding to different alleles of CBX4. Simultaneously, we validated the findings of population-based cross-sectional studies using computer models, in vitro cell assays and animal trials. The number of publications on hearing loss susceptibility genes and SNPs has increased steadily, from 42 in 2009 to 215 in 2021. NIHL-associated susceptibility genes are implicated in heat shock protein (21), oxidative stress (22), Notch signaling (23) and apoptotic signaling pathways (24), but only a few investigations about the PTM signaling pathway.

Noise has been demonstrated to induce hypoxia and disturbances in the energy metabolism of cochlear hair cells, ultimately resulting in hair cell death. Hair cell death occurs through two different pathways: apoptosis and necrosis, with apoptosis being the more prevalent (25). Increased oxygen free radicals in the mitochondria of cochlear hair cells induce DNA damage, which results in the release of apoptosis-inducing factors into the cytoplasm, activating caspase-3 and initiating cell apoptosis (26). PTMs are critical regulators of cell biology since they participate in processes such as the quick on/off switching of signaling networks and the regulation of enzyme activity (11). PTMs like glycosylation, acetylation, ubiquitination, methylation and SUMOylation at particular regions in proteins play critical roles in the development and function of the cochlea (12). As a result, their mutation or silencing may lead to deafness. Research on the role of SUMOylation as an important PTMs in the cochlea is still in a preliminary stage, and little research data on hearing loss are available.

First, in this cross-sectional study, we examined the genetic associations of 31 SUMOylation (SAE1, CBX4, RANBP2, UBE2I and UBA2) SNPs in 1170 noise-exposed subjects susceptibility to NIHL. CBX4 rs1285249 and rs1285250 may be associated with NIHL susceptibility. Notably, the T allele frequency was significantly higher in cases than in controls, suggesting that the T allele was associated with an elevated risk of NIHL and that the TT genotype was a risk genotype. Subsequent haplotype analysis showed that the haplotype ATGT (rs4889898–rs1285243–rs1285249–rs1285250) increasing NIHL susceptibility. To explore potential biological pathways underlying known CBX4 rs1285249 and rs1285250, we conducted analyses using HaploReg and Luciferase assay. The rs1285250 allele has the highest binding score to SP1, the T allele has the strongest binding ability, and the luciferase activity was substantially different from the matched C allele. SP1 was the first mammalian transcription factor to be described and remains one of the most extensively investigated. It acts as a transcription factor for a large number of genes, including tissue-specific genes, cell cycle regulators and housekeeping genes. A growing body of evidence indicates that ROS regulate SUMOylation, both during redox signaling and under conditions of severe oxidative stress. Notably, SUMOylation plays a critical function in regulating ROS homeostasis by regulating ROS generation and clearance (15,27). By developing a mouse model of NIHL, we found that noise exposure can lead to increased levels of oxidative stress. This result is similar to those obtained by others (28,29). Meanwhile, noise-exposed mice had dramatically activated SP1/CBX4 signaling pathways, resulting in decreased anti-apoptotic protein expression and increased apoptotic protein expression. To demonstrate that SP1 can act as transcriptional enhancers, we used Sp1 siRNA to knockdown endogenous Sp1 expression (si-Sp1). Expression levels of genes in Cbx4 decreased upon SP1 knockdown in HEI-OC1 cells. Furthermore, the rate of apoptosis was significantly lower in the SP1-knockdown group than in the si-NC transfection group. The results presented previously show that the SP1 transcription factor may play the role of enhancer to boost the expression of the Cbx4 gene and promote cell apoptosis. The results of CBX4 promoting cell apoptosis were similar to those in the study by Peuget et al. (30). CBX4 mediates SUMOylation of TP53INP1 in response to DNA damage, thereby favoring P53 transcriptional activity on pro-apoptotic genes.

Nevertheless, our research has some limitations. (i) In comparison with previous research, our survey involved a large population. However, owing to the insignificant biological consequences of a single SNP, the sample size may have been insufficient for an effective statistical analysis. As a result, more research with a larger sample size is necessary to confirm the influence of CBX4 polymorphisms on the risk of NIHL. (ii) The case–control study’s participants were exclusively Chinese Han. As a consequence, our findings may be limited to the Chinese Han population. (iii) Consistent with previous studies (31,32), the current study focused on male workers and male mice and did not adequately consider the effects of noise on hearing loss in females (33). There are still many female workers in China who are exposed in the workplace to loud noise, but, sadly, the conclusions drawn from this designed study may only apply to men.

This research establishes the first evidence that the CBX4 gene may serve as a valuable biomarker for NIHL genetic susceptibility. Meanwhile, it is possible that the interplay of genetic variation and noise exposure plays a significant role in the pathogenesis of NIHL. Furthermore, the SP1 transcription factor may act as an enhancer by binding the rs1285250 T/C allele with different strengths to promote the expression of the CBX4 gene. By exploring CBX4 association with NIHL, it is possible to improve our understanding of the biological effects of SUMOylation may pave the way for new approaches to preventing and treating hearing loss and deafness. In addition, key information on how PTMs alter gene regulatory networks during ototoxic damage may open up a new and promising therapeutic avenue.

Materials and Methods

Study subjects

Overall, 1260 subjects from four companies were involved in the machinery manufacturing corporation, chemical fiber company and machines. Trained personnel interviewed the participants and filled out a structured questionnaire, which included questions about age, sex, employment duration, tobacco, alcohol consumption, history of occupation and disease history (for further details, see the Supplementary material). Overall, 1170 subjects completed the interviews (response rate, 92.86%). The inclusion criteria for subjects exposed to benzene included (i) working in a workplace with exposure to noise > 1 year; (ii) a permanent employment contract; (iii) signed consent forms; (iv) environmental exposure to noise levels above 85 dB (A) and (v) ethnic Han Chinese ethnicity. The exclusion criteria: (i) history of head injury, otological illness or other conditions that affect hearing loss; (ii) previous or current ototoxic medical treatment and/or possibly detrimental noise exposure while serving in the military; (iii) history of recreational noise exposure and (iv) incompletely filled questionnaires and monitoring data on occupational health. All participants worked at least 5 days a week, 8 h a day, with at least 2 h of exposure to the noise environment.

Environmental noise monitoring

Noise exposure levels were measured using sound pressure individual noise meters (Quest NoisePro DL Dosimeter, Mara Industrial ID: 144 147-169 778) carried by noise-exposed workers throughout their working time three times a year, according to the classification of occupational hazards in the workplace: noise (GBZ/T 229.4–2012). Meanwhile, a sound pressure noise meter was used to measure it at 10 a.m., 3 p.m. and 5 p.m. Lex recorded the result for 8 h to assess the noise exposure level.

Hearing assessment and definition of NIHL

According to the Chinese Diagnostic Criteria of Occupational NIHL, the hearing thresholds of the participant’s left and right ears were evaluated at 500, 1000, 2000, 3000, 4000 and 6000 Hz (GBZ49-2014). The tests were conducted with a Madsen electronics voyager 522 portable diagnostic audiometer in a soundproof environment with a background noise level of 25 dB (A) (Madsen electronics, Taastrup, Denmark, S/N: 72 189). Before the inspection, all of the participants had to be out of the noise environment for at least 12 h. According to WHO criteria, those with a high binaural frequency (3000, 4000 and 6000 Hz) average hearing threshold > 25 dB (A) was classed as NIHL, whereas those with hearing thresholds < 25 dB in both ears were classified as controls.

SNP selection

Han Chinese genotype data from the 1000 Genome Project were used to search for possible SNPs and then select the most promising candidates. The tagSNPs were found using the Haploview 4.2 program with the pairwise option and a threshold of r² = 0.8. We further examined the minor allele frequencies (MAFs) of these tagSNPs in the NCBI SNP database, excluding those with no reports in the population or MAFs of < 0.05. The final candidate tagSNPs, relative genomic locations and linkage disequilibrium (LD) are represented in Figure 1.

DNA extraction and genotyping

Each participant’s peripheral blood was collected and put into an ACD tube in the amount of 10 ml. Proteinase K digestion and phenol-chloroform extraction were used to extract genomic DNA from peripheral blood samples. Using sterile procedures, DNA was extracted in the laminar flow hood. The SNaPshot was used to identify the genotypes of the 31 tagSNPs. SNaPshot is a single-base extension (SBE) assay with capillary electrophoresis as its detection system. This multiplexing technique offers the advantage of easy integration into laboratories without the requirement for any additional equipment. Furthermore, the SNP panels from SNaPshot assays can be incorporated into customized panels for massively parallel sequencing (34,35). Shanghai Biowing Applied Biotechnology Co., Ltd company used multiplex PCR and next-generation sequencing to identify the genotype (36). In this investigation, the genotyping method was done under double-blind conditions. About 10% of all samples were chosen at random for triplicate tests, and the findings were 100% consistent.

Animal experiments

All experiments were performed on male C57BL/6 (6–8 weeks) mice. Beijing Viton Lever Laboratory Animal Technology Co., Ltd provided the animals, which were kept on a 12 h light/dark cycle in relative humidity and controlled temperatures with a background noise level (<60 dB). Water and food were freely available to the mice at all times. For each mouse, one cochlea was collected for immunohistochemistry and the other for mRNA and protein extraction. A total of 12 mice were randomly separated into two groups: control and noise-exposed. Aside from the noise exposure settings, both groups received the same treatment. All animal experiments were authorized by Southeast University’s Animal Ethics Committee and conducted in accordance with the Nanjing Committee for the Use and Care of Laboratory Animals.

Noise exposure

Noise exposure groups were exposed to an octave band (8–16 kHz) noise at 120 dB SPL for 2 h per day on 2 consecutive days. The noise exposure model was established as previously reported (37) with some modifications.

Auditory brainstem response

ABR tests were performed as previously reported utilizing a Tucker-Davis (TDT) BioSig System III (Tucker-Davis Technologies, Alachua, FL, USA). Following anesthesia, mice were subcutaneously implanted with subdermal needle electrodes at the vertex and below the left and right ears. The ABR was tested at four different frequencies: 4, 8, 16, 24 and 32 kHz. Auditory thresholds were determined by decreasing the sound level of each stimulus in 5 dB increments until the response had the lowest sound intensity with repeatable and identifiable waves.

Tissue preparation

Mice injected with sedatives were decapitated after ABR testing. The cochlea was obtained and decalcified in 4% sodium ethylenediaminetetraacetic acid (EDTA) for 2 days after being fixed with 4% PFA at 4°C overnight. Cochlear tissues were excised and snap-frozen in liquid nitrogen before being kept at −80°C for RNA and protein extraction.

Immunofluorescence

The cochlea was soaked in 4% PFA and fixed on a shaker at room temperature for 1 h, and then it was decalcified in 0.12 M EDTA for 2 days at room temperature. After three times washing with PBS, it was placed in PBS solution containing 5% donkey serum, 0.5% Triton X100, 0.02% sodium azide and 1% bovine serum albumin and sealed for 1 h at room temperature. It was treated with PBT-1 (PBS solution containing 2.5% donkey serum, 0.1% Triton X100, 0.02% sodium azide and 1% bovine serum albumin) containing primary antibody (Myosin7a, 1: 1000, Proteus Bioscience, #25–6790) at 4°C overnight. After three times washes with PBST, PBT-2 containing secondary antibody (PBS solution containing 0.1% Triton X100 and 1% bovine serum albumin) was used for incubation at room temperature for 1 h. Each sample received 8 μl Dako and a coat of nail polish to seal the slice. An LSM700 laser confocal microscope was used to take the images (Carl Zeiss AG, German).

Cell culture

Dulbeco’s modified eagle medium (DMEM; Gibco-BRL, USA) was used to cultivate cochlear HEI-OC1 cells (kindly donated by F. Kalinec at the House Ear Institute in Los Angeles, CA, USA) at 33°C in a 10% CO₂ condition in the presence of 10% fetal bovine serum (Cat #. 10099141C). The Shanghai Cell Bank of the Chinese Academy of Sciences (Shanghai, China) provided HEK-293 cell lines, which were cultured in DMEM supplemented with 10% FBS at 37°C under 5% CO₂.

Cell transfection

Guangzhou RiboBio developed and manufactured Sp1 siRNA (si-sp1, GGATGGTTCTGGTCAAATA; China). Three siRNAs targeting Sp1 were developed and produced, with the most efficient siRNA (si-Sp1) being selected in the following studies based on quantitative real-time (qPCR) results. For cell transfection, the Lipofectamine 2000 system (Invitrogen, USA) was used according to the manufacturer’s instructions. We replenished the media with a new complete culture medium 4–6 h after transfection. The HEI-OC1 cells were harvested 48 h after transfection for immediate usage.

Prediction of the binding of the putative to the CBX4 polymorphism

The transcription factor that may target the CBX4 intron variation (rs1285249 and rs1285250) was predicted using the SNPinfo Web Server’s SNP Function Prediction (FuncPred) software. SP1 was identified as a potential transcription factor that targets CBX4’s intron variant.

Luciferase assay

The rs1285250 T/C region fragment was used as the core to synthesize a double-stranded oligonucleotide with a sequence length of 21 bp. The oligonucleotides were then cloned into the pGL3-promoter vector, which contains a simian virus 40 (SV40) promoter (Promega, Madison, WI, USA). Lipofectamine 2000 (Cat #. 11 668 019) was used to transfect HEK293 cells with 100 ng reporter constructs and 10 ng pRL-TK Renilla luciferase vector (Promega) according to the manufacturer’s protocol. The effectiveness of transfection as compared with the activity of Renilla luciferase. Eighteen hours after transfection, the medium was transferred to the growth medium. The luciferase activity of the cells was evaluated with the Dual-Luciferase Reporter Assay System 48 h after the medium was changed (Promega).

Total RNA extraction and qPCR

To collect high-quality total RNA, researchers used TRIzol Reagent (Gibco, USA), which was then mixed with Invitrogen SuperScript III Reverse Transcriptase to produce cDNA. SYBR Green qPCR MasterMix and an Applied Biosystems StepOne qPCR equipment were used in the qPCR tests (Carlsbad, CA, USA). The levels of β-actin expression served as internal controls for mRNA. Supplementary Material, Table S2 illustrates the primers for the candidate genes.

Western blot assay

Total proteins were separated by SDS/PAGE on a 10–12.5% gel and then transferred onto polyvinylidene fluoride (PVDF) membranes after cells were lysed in RIPA lysis buffer (Sigma-Aldrich, USA). After being blocked, the membranes were incubated at 4°C overnight with the prescribed dilutions of specific primary antibodies. HRP-conjugated secondary antibodies and an ECL system were used to detect the specific protein bands (Amersham Biosciences, UK). At dilutions of 1: 1000, primary antibodies against CBX4, SP1, CASP3, A-CASP3, BCL-2, BAX and β-ACTIN were utilized.

Apoptosis assay

Flow cytometry was used to investigate the apoptotic rate using the FITC Annexin V Apoptosis Detection Kit I (BD Biosciences, USA). Cells that had been transfected with si-Sp1 for 24 h were washed in PBS before being resuspended in a binding buffer containing FITC-labeled Annexin V and PI. The cells were cultured for 15 min in a dark chamber before being analyzed using flow cytometry.

Measurement of malondialdehyde, glutathione peroxidase and superoxide dismutase on serum

Serum was isolated via the retro-orbital vein from anesthetized mice. AndyGene Biotechnology Co. provided the malondialdehyde (MDA; Cat #. AD11335Hu), glutathione peroxidase (GSH-Px; Cat #. AD12650Hu) and superoxide dismutase (SOD; Cat #. AD12043Hu) ELISA kits (Beijing, China). ELISA was used to assess the levels of MDA, GSH-Px and SOD in the serum. At a wavelength of 450 nm, the absorbance was measured using a microplate reader.

Statistical analyses

The normal distribution of all data was examined using the Shapiro–Wilk test. The Student’s t-test or the χ²-test were used to compare demographic and genotype data for NIHL patients and controls. If normality was rejected by Shapiro–Wilk test, Mann–Whitney U-test was applied for non-parametric data. Declaration of the categorical variables was carried out as frequencies to achieve the analysis of the statistical significance of categorical variables (%). A goodness-of-fit 2-test was used to determine if genotypes were in HWE. The odds ratio (OR) and 95% confidence intervals (CI) for the association between the genotypes of polymorphisms and the risk of NIHL were calculated using logistic regression analysis with adjustments for age, gender, cigarette smoking and alcohol consumption. D′ and R2 were used to determine LD between polymorphisms, and Haploview 4.1 software was used to illustrate the characterization of these patterns. Multiple comparisons corrections were carried out with Sidak, Holm’s correction. Using SPSS 22.0 software, all tests were two-sided, and P < 0.05 was considered statistically significant.

Data Availability

All data generated or analyzed during this study are included in this published article (and its supplementary information files).

Acknowledgements

We thank all workers for participating in this study.

Conflict of Interest statement. The authors declare that they have no competing interests.

Funding

Key Program of Jiangsu Provincial Health and Health Commission Medical Research (K2019026); Jiangsu Province’s Outstanding Medical Academic Leader program (CXTDA2017029).

Consent for publication

Consent to publish was obtained from all participants.

Author’s statement

This is an original article and presents recent work from our laboratory. The data have been neither published nor submitted for publication elsewhere in any language. Each author has contributed to the article and can take full responsibility for its contents.

Authors’ contributions

Boshen Wang, Liu Wan and Juan Zhang designed the study. Juan Zhang and Baoli Zhu supervised the study. Peng Sun, Liu Wan and Ludi Zhang jointly performed the experiments. Boshen Wang, Hengdong Zhang and Baoli Zhu contributed to the collection and analysis of the workers’ biological samples. Ludi Zhang and Liu Wan helped with the animal study. Yuepu Pu, Boshen Wang and Liu Wan interpreted the data and drafted the manuscript. All authors read and approved the final manuscript.

Author notes

Boshen Wang and Liu Wan contributed equally to this article.