30 male SPF Wistar rats (body weight: 160-180 g; age，4–5 weeks old) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). The animal license was NO. SCXK (jing) 2016–0006. The rats were housed with controlled light (12 h light/dark cycle), constant temperature (20–26°C), and ambient humidity (40–70%) before being used for experiments (Experimental Animal Center, Tianjin Institute of Health and Environmental Medicine, China). Based on the current upper limit of the safety allowance recommended by the U.S. Occupational Safety and Health Administration (OSHA),¹ we chose a noise level of 95 dB sound pressure level (SPL) for 4 h per day (4 h/d). To investigate the impact differences between noise exposure of different intensities, 105 dB SPL for 4 h/d was also used for comparison. As a rat was found to have middle ear infection before the experiment, it was excluded and only 29 rats were included in this study. After 1 week of acclimatization, the rats were numbered according to their body weight and the ones with similar weight were then randomly divided into three groups: Ctr group (exposed to background noise ≤40 dB SPL, n = 9), LN group (exposed to 95 dB SPL white noise, n = 10) and HN group (exposed to 105 dB SPL white noise, n = 10). As reported by Arifin and Zahiruddin (2017), minimum sample sizes for ‘group comparison, one-way-analysis of variance’ design were five animals per group to keep degrees of freedom within the range of 10 to 20. Therefore, we met the sample size required for the experiment with nine or ten rats per group. For the rats in LN group and HN group, exposure was for 4 h/d at 8:30–12:30 for 30 days, referred to a previous study (Cui et al., 2012). In Brief, white noise was generated by a generator (BK 3560C; Brüel & Kjær Instruments, Nærum, Denmark). The main spectrum of the noise signal from the generator was in the range of 20–20,000 Hz, amplified by a power amplifier (Yong-Sheng Audio P-150D; The Third Institute of China Electronic Technology Group, Beijing, China), and emitted from a loudspeaker (ZM-16S; Tianjin Zenmay Electroacoustic Equipment Co., Tianjin, China). To exclude confounding factors, food and water in the Ctr group were also the same as the LN and HN groups at the same time.

The experimental proposals and procedures for the care and treatment of the rats were carried out according to the ARRIVE guidelines (Percie du Sert et al., 2020), and were approved by the Institutional Animal Use and Care Committee, Tianjin Institute of Health and Environmental Medicine (Authorization No. IACUC of AMMS-04-2020-063).

ABR thresholds of rats before and after noise exposure were measured with auditory evoked potential and otoacoustic emission instrument Smart-EP ASSR (Smart-EP Manual Version 3.97 test system, Smart Listening, United States). All the animals were mildly anesthetized (1% Pentobarbital sodium, 70 mg/kg) and placed in the anechoic room. According to the instructions and previous studies (Paciello et al., 2021, 2023), click stimuli and tone bursts of pure tones from 2 k-32 k HZ (Filter bandwidth: 100 ~ 16,000 HZ, stimulation frequency: 21.10/s, scanning time: 12.8 ms, superimposed 1,024 times) were used and the ABR thresholds of the ears were measured. For the detection, high frequency headphones were used to produce the 16 k and 32 k HZ sounds while the ER3 earphone was used to give the 2 k, 4 k and 8 k HZ sounds. The stimulation starts at 90 dB SPL and decreases at 10 dB SPL. When a threshold occurs, the measurement is repeated twice at the threshold of ±5 dB. The ABR thresholds of the rats among different groups were compared with Kruskal-Wallis test and p < 0.05 was considered significant.

After the last noise exposure, the rats were fasted for 12 h. The blood was collected from the abdominal aorta of anesthetized rats (1% pentobarbital sodium, 0.7 mg/10 g body weight) and centrifuged at 3000 rpm for 10 min at low temperature. The serum was collected, quenched in liquid nitrogen, and stored at −80°C for serum biochemical detection and metabolomic analysis. To avoid contamination, fresh feces (5 g) were collected from the rectal end of the rats and stored in frozen storage tubes after liquid nitrogen quenching for 16S rDNA gene sequencing.

The deep anesthetized rats above were then sacrificed by cervical dislocation. The skull was immediately dissected and the cochlea was obtained. The cochlea tissues were pre-cooled with 4% paraformaldehyde and 2.5% glutaraldehyde solution, and then fixed overnight at 4°C. With the labyrinth parts exposed, the specimens were put into 10% ethylenediaminetetraacetic acid (EDTA) for decalcification. When the cochleae were translucent, they were stored in fixative solution at 4°C for subsequent detection.

The decalcified cochlear tissues were subjected to gradient dehydration, transparency, wax immersion, embedding, serial sectioning, and Hematoxylin–Eosin (HE) staining as described in previous studies (Zhao et al., 2021; Li et al., 2023). With light microscopic observation, the morphological features of the cochlear tissues in different groups were visualized and compared. Laser confocal microscopy (Leica TCS SP8, Leica Microsystems, Germany) was also used to observe the morphological changes of the cochleae.

To investigate the effects of noise exposure on liver function, and regulations of glucose and lipids, the serum levels of alanine transaminase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), γ-glutamine transferase (γ-GT), total gallbladder juice acids (TBIL), direct bile acids (DBIL), glucose (GLU), glycosylated serum protein (GSP), triglycerides (TG), total cholesterol (CHO), high-density lipoprotein (HDL), low-density lipoprotein (LDL), creatinine (CREA), UREA and uric acid (UA) were detected with full-automatic biochemical analyzer (Catalyst One® Chemistry Analyzer, IDEXX Laboratories Inc., United States) according to the instructions. Kruskal-Wallis test was used for comparisons among different groups and p < 0.05 was considered significant.

Untargeted metabolomics tests were performed. Due to clotting during sample acquisition, four samples were excluded and eight samples from Ctr group, seven from LN group and eight from HN group were included for analysis. Each 100 μL serum sample was mixed with 400 μL of pre-cooled methanol/acetonitrile (1: 1, v/v) and then centrifuged for 15 min with 14,000 rpm at 4°C. The supernatant was dried in a vacuum centrifuge and hydrop interaction liquid chromatography (HILIC) separation was performed using a 2.1 mm × 100 mm ACQUIY UPLC BEH 1.7 μm column (waters, Ireland). Detailed processes of extraction, detection, and quality control measures for metabolites in UPLC/Q-TOF-MS analysis are described in previous studies (Liu et al., 2020; Zhang X. et al., 2022).

Raw MS data were subjected to peak extraction to obtain information on characteristic peaks such as m/z values. Metabolites were identified by comparison with the database. Finally, the data files were subjected to univariate statistical analysis (Student’s t-test and fold change analysis) and multivariate statistical analysis-principal component analysis (PCA) and orthogonal partial least squares discrimination analysis (OPLS-DA) using the “ropls” package in R3.6.1. The variable importance in the projection (VIP) value of each variable in the prediction was calculated by the OPLS-DA model. Significantly different metabolites were identified with the thresholds of VIP > 1, p < 0.05 and Fold change (FC) > 1.5 or FC < 0.67.

16S rDNA gene sequencing on rat feces were performed and the differential flora were screened. As some samples were insufficient or failed DNA extraction, only seven samples from Ctr group, six samples from LN group and nine samples from HN group were included for sequencing analysis. DNA was extracted from fecal samples according to the instructions in the fecal DNA extraction kit (E.Z.N.A. ®Stool DNA Kit, D4015, Omega, Inc., United States). Polymerase chain reaction (PCR) products were confirmed by 2% agarose gel electrophoresis and PCR products. DNA was purified with AMPure XT beads (Beckman Coulter Genomics, Danvers, MA, USA) and quantified using Qubit (Invitrogen, USA). PCR amplification of the V3-V4 region of 16S rDNA from fecal samples was performed using primers 341F (5’-CCTACGGGNGGCWGCAG-3′) and 805R (5’-GACTACHVGGGTATCTAATCC-3′). Equal amounts of amplicons were mixed for sequencing. The size and number of amplicon libraries were evaluated on Agilent 2100 Bioanalyzer (Agilent, United States) and Illumina library quantification kit (Kapa Biosciences, Woburn, MA, United States). Libraries were sequenced on the NovaSeq PE250 platform.

Feature tables and feature sequences were obtained by filtering the raw data according to fqtrim (v0.94). Alpha diversity and beta diversity were randomly normalized to the same sequence. The feature abundance was then normalized according to the SILVA (release: 132) classifier. Species diversity and complexity of the samples were analyzed. The alpha diversity indexes including Chao1 index and Shannon index were evaluated with “vegan” R package. Beta Diversity were calculated with the “QIIME2” R package (v3.5.2). Sequence comparisons were performed using the blast method. Feature sequences were annotated with each representative sequence using the SILVA database. Finally, genera with linear discriminant analysis (LDA) values >3 were identified as significantly differential through linear discriminant analysis effect size (LEfSe) analysis. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (level3) associated with the gut microbiota abundances in different groups were predicated with PICRUSt.² The scores of the pathways among the three groups were compared with Kruskal-Wallis test and p < 0.05 was considered significant.

To ensure a one-to-one correspondence between metabolites and gut microbes, we analyzed only the omics data of the rats underwent both metabolomic and microbiomic assays for the combined analyses. Only the significant differential bacterial microbiota (LDA value >3 and p < 0.05) and metabolites (VIP > 1, p < 0.05, and FC < 0.67 or FC > 1.5) were included. Spearman correlation analysis was performed in R and p < 0.05 was considered significant. Clustering correlations were performed using the OmicStudio tools at https://www.omicstudio.cn.

As shown in Figure 1, at all the frequencies, the ABR thresholds of the left and right ears of the Ctr rats presented to be the lowest among the three groups, indicating the significant increase of ABR thresholds of the noise-exposed rats after noise exposure. In contrast, HN rats were shown to have the highest level of ABR thresholds among the three groups. There were increase trends of ABR threshold displacements at click, 4 k, 8 k, and 16 k HZ were also indicated in Ctr, LN, and HN rats. These results indicated the NIHL of the noise-exposed rats and the stronger impacts of higher-intensity noise.

There were obvious cochlea injuries in the noise-exposed rats. Through HE-staining (Figure 2), it was presented that, comparing with the rats in Ctr group (Figures 2A,B), the spiral ganglion cells in the LN group (Figures 2C,D) decreased with the nucleus swelled and enlarged, and the ratio of cytoplasm and nucleus imbalanced. For the HN group (Figures 2E,F), the spiral ganglion cells presented loose structures, a significant decrease of the nerve cell number, and a disorganized nucleoplasm ratio. Furthermore, in the HN group, a large number of swollen nuclei were also shown and the cell boundaries were obscure. For a few cells, their nucleoli and nuclear membranes also disappeared. These injuries might account for the NIHL in the noise-exposed rats and indicated the stronger impacts of noise-exposure of higher intensity.

Confocal laser microscopy observation presented similar results. As shown in Figure 3. In the Ctr group (Figure 3A), the cells were arranged neatly. In contrast, external hair cell deletion and nuclear solidification were shown in the LN group (Figure 3B). For the HN group (Figure 3C), the arrangement of the external hair cells was disordered and most of cells presented nuclear displacement, nuclear solidification, nuclear fragmentation, and nuclear lysis, indicating the death and deletion of the external hair cells. However, no significant deletions and misalignment of internal auditory hair cells were shown in the three groups.

As shown in Figures 4A–C, the levels of ALT, AST, and γ-GT were higher in the LN and HN groups that in the control group. Furthermore, the HN presented highest ALT and AST levels among the three groups. Although no significant TBIL was shown among the three groups (Figure 4E), the DBIL were also shown to be higher in the noise-exposed LN and HN rats than the controls (Figure 4D). These results indicated the potential liver injuries of the rats due to noise-exposure, especially in the HN group. The levels of serum glucose and glycosylated serum protein (GSP) also appeared to be dysregulated in the noise-exposed rats (Figures 4F,G), indicating the impact of noise exposure on glucose metabolism. The levels of blood lipids were also affected by the noise-exposure. As shown in Figures 4H,I, TG and LDL increase were shown in noise-exposed rats. Considering the associations of LDL and TG with atherosclerosis (AS) (Yan et al., 2022; Yu et al., 2022), it was indicated that higher noise might be a risk factor of AS. As no significant difference of CHO and HDL was shown among the three groups (Figures 4J,K), the selectiveness of the effects of noise exposure on blood lipids was indicated. Interestingly, no significant difference of CREA, UREA and UA were shown between different groups (Figures 4L–N).

To understand the changes in the metabolic profiles of rat serum after noise exposure, we performed a metabolomic analysis. Pearson correlation analysis on quality control (QC) samples presented a correlation coefficient of R > 0.9 in the negative (Supplementary Figure S1A) and the positive (Supplementary Figure S1B) mode, indicating good reproducibility of the experiments. The metabolites and their levels in different groups were shown in Supplementary Tables S1, S2.

Through OPLS-DA (Supplementary Figure S2), each two of the three groups could be discriminated clearly, indicating the significant differences of metabolism between the three groups. Through Wilcoxon test, the metabolic differences between each two of the three groups were investigated. With the threshold of VIP >1 and p < 0.05, the differential metabolites were identified (Figure 5). There were 36 (up-regulated: 15, down-regulated: 21) (Figure 5A) and 62 (up-regulated: 27, down-regulated: 35) (Figure 5B) differential metabolites between HN and Ctr groups in the negative and positive modes, respectively. Comparing with the control samples, 34 metabolites (up-regulated: 8, down-regulated: 26) (Figure 5C) and 36 metabolites (up-regulated: 12, down-regulated: 24) (Figure 5D) were found to be dysregulated in LN samples in the negative mode and positive mode, respectively. There were common dysregulated metabolites in the LN and HN groups. In fact, four metabolites (1-oleoyl-2-palmitoyl-sn-glycero-3-phosphocholine, 3-hydroxybutyric acid, Pi 38:4 and Pe 38:4) were upregulated while seven metabolites (Indolelactic acid, Hippuric acid, 2,6-dihydroxybenzoic acid, 7-keto-3.alpha.,12-.alpha.-dihydroxycholanic acid, Acetaminophen sulfate, Isatin, and Quillaic acid) were shown to be downregulated in the two noise-exposed groups. Furthermore, comparing with LN samples, 20 metabolites (negative mode: 13, positive mode: 7) were higher while 11 metabolites (negative mode: 7, positive mode: 4) were lower in HN samples (Figures 5E,F). The common dysregulated metabolites were also provided in the Supplementary Tables (Fig 5_information_supple).

Through KEGG pathway enrichment analyses, the dysregulated metabolites in HN and LN groups were shown to be associated with 30 and 12 biological processes, respectively (Figures 6A,B). Noticeably, there were seven pathways (Glycosaminoglycan biosynthesis - keratan sulfate, N-Glycan biosynthesis, Adipocytokine signaling pathway, Peroxisome, Bile secretion, Various types of N-glycan biosynthesis, and cAMP signaling pathway) consistently dysregulated due to noise-exposure in the HN and LN groups while the other processes were different (Figure 6C).

The 16 s rDNA sequencing data was deposited in the NCBI Sequence Read Archive (SRA, http://www.ncbi.nlm.nih.gov/sra), with the accession number PRJNA967148. Through 16S rDNA gene sequencing analyses, a total of 14,358 operational taxonomic units (OTUs) were identified. As shown in Supplementary Figure S3, there were 1,423 common OTUs between Ctr and LN groups, 1,470 common OTUs between Ctr and HN groups, 1,494 common OTUs between LN and HN groups, and 1,032 common OTUs in the three groups. In contrast, the differences among the three groups were also shown and there were 4,025, 3,239, and 4,731 specific OTUs for Ctr group, HN group, and LN group, respectively. The group-specific OTUs indicated the impact of noise exposure of different intensity on the gut microbiome of the hosts.

Through Kruskal-Wallis test, the Chao1 index presented significant difference among the three groups (Figure 7A), indicating that different noise exposure with different levels might have different effects on the host gut microbiota. In Figure 7B, the Shannon index of HN group was shown to have a lower trend to be the lowest among the three groups, indicating the decrease trend of microbiota diversity (abundance and richness) in HN group. In Figure 7C, PCA analysis was shown to be able be discriminate the three groups, indicating the significant difference of the ß diversity in the three groups.

Through LEfSe analysis, the microbiotas with significant abundance differences among different groups were identified. As shown in Figure 8A, the red-, green-, and blue-labeled microbial species were shown to be enriched in the control group, HN group, and LN group, respectively. The potential biomarkers with LDA values >3 was also shown Figures 8B. These results indicated the complex impacts of noise exposure on gut microbiota in the host. Comparing with the Ctr group, there were 14 and six significant microbial genera with LDA value >3 and p < 0.05 in HN group (Figure 8C) and LN group (Figure 8D), respectively. It was indicated that noise exposure of different intensity might lead to different impacts on gut microbiotas of the host.

The scores 283 KEGG pathways were predicated (Supplementary Table S3) and compared between the three groups. There were 49.1% (139/283) of the pathways were shown to be dysregulated after noise exposure (Supplementary Table S4). As shown in Figure 9A, the scores of D-arginine and D-ornithine metabolism were shown to be lowest in LN rats among the three groups. For many pathways, highest scores were shown in the HN rats. The representative pathways were shown in Figures 9B–F. Figure 9B presented highest scores of DNA repair and recombination proteins in HN rats, indicating the increase of DNA repair activity due to noise exposure. The highest scores of primary immunodeficiency in HN group (Figure 9C) indicated the potential damage by high-intensity noise exposure to the immunity of exposers. The highest activity of lipid metabolism (Figure 9D) was consistently with the increased TG level in the HN group. The highest scores of type I (Figure 9E) and type II (Figure 9F) diabetes mellitus was consistent the dysregulated serum of GLU and GSP levels in the noise-exposed rats above, implicating the impacts of noise exposure on the GLU regulation.

The associations between the relative abundance of significantly dysregulated genera above and the expression of significantly dysregulated metabolites were analyzed. In the HN group (Figure 10A), Corynebacterium_1 and Prevotellaceae_UCG_001 were shown to be significantly correlated with 23 and 25 metabolites, respectively. Erysipelotrichaceae_UCG_003, Frisingicoccus, and Peptococcaceae_unclassified were significantly correlated with 20, 14, and 21 metabolites, respectively. In the LN group (Figure 10B), Romboutsia and Turicibacter were significantly associated with 21 (20 positive and 1 negative correlation), and 19 (18 positive and 1 negative correlation) metabolites, respectively. While Eubacterium_ruminantium_group was negatively correlated with 12 metabolites and two positively correlated. These results suggest close relations between the gut microbiota and the metabolic profiles of the rats during noise exposure.