Characterization of the Middle Ear Microbiome in Meniere's Disease by 16S rRNA Gene Sequencing Compared to Chronic Suppurative Otitis Media Authors: Yu‐Ting Tao, Lin Wang, Hong‐wei Zheng, Li‐Sheng Yu Published: 2026 ABSTRACT ABSTRACTObjectiveTo investigate the middle ear microbiome in Meniere's disease (MD) patients.MethodsMiddle ear smears from 10 MD patients (ME group) and 10 chronic suppurative otitis patients (OE group) were collected during surgery and processed via 16S rRNA gene sequencing.ResultsThe ME group demonstrated significant differences in alpha diversity (Shannon/Simpson index, p < 0.01) and beta diversity (ANOSIM/PERMANOVA, the Bray‐Curtis distance matrix, R = 0.144, p < 0.05; R 2 = 0.107, p < 0.05) compared to OE group, and the PERMDISP showed lower heterogeneity in the ME group (p < 0.01). Ruminococcus, Ochrobactrum, and Halomonas at the genus level were enriched in the ME group, and the difference reached marginal significance. The linear discriminant analysis effect size analysis (LefSe) identified Bacteroides, Halomonas, and Prevotella as significantly enriched in ME, while Neisseriales was downregulated (|LDA| > 2, p < 0.05). BugBase phenotype analysis revealed that facultatively anaerobic bacteria and bacteria containing mobile genetic elements were significantly reduced in the ME group compared to the OE group (p < 0.05). The proportion of anaerobes was higher in the ME group (p = 0.123). Nine KEGG metabolic pathways were enriched in the ME group, though the difference was not significant after correction.ConclusionThese findings suggested that inner ear inflammation resulted from middle ear dysbiosis may contribute to MD pathogenesis, offering novel perspectives for MD etiology and therapies. Introduction Meniere's Disease (MD) is an idiopathic inner ear disorder defined by spontaneous episodes of vertigo, fluctuating sensorineural hearing loss, tinnitus, and aural fullness [1]. These debilitating symptoms can progress to severe and bilateral involvement over time, causing physical and mental burden and impairing the quality of life [2]. While endolymphatic hydrops (the excessive accumulation of endolymph) is widely acknowledged as the crucial histopathological feature of MD [3], the pathophysiology remained elusive. Genetic abnormality and anatomical variation can result in susceptibility to MD, and environmental factors, including noise [4], infection (bacterial [5] or viral [6]), immune dysregulation [7], allergy and trauma [8, 9], may contribute to episodic symptoms. Bacterial infection has been a suspected trigger of MD since Paparella MM reported in 1983 that MD symptoms could occur in patients of chronic otitis media concurrently or years later [5]. A cohort study reported childhood chronic suppurative otitis media (CSOM) correlated with increased risk of dizziness in adulthood, suggesting the persistent inner ear damage from previous middle ear infection [10]. Preliminary study revealed that membranous occlusive structures and fibrosis were frequently observed in the antrum and mastoid process of MD, and enlargement of the antrum with the thorough removal of mastoid cells could achieve long‐term alleviation of vertigo recurrence [11]. These findings suggest a potential involvement of the middle ear microenvironment in the pathogenesis of MD. Molecular technologies have significantly enhanced the sensitivity of microbial detection in recent years, and bacterial communities in healthy middle ears, CSOM and acquired cholesteatoma have been unveiled via 16S rRNA gene sequencing [12, 13]. However, the characteristics of the middle ear microbiome (MEM) in MD patients remain unexplored. In this study, the middle ear microbial profiles in MD patients and the positive control (CSOM patients) were revealed via 16S rRNA gene sequencing. Through this research, we expect to provide novel microbiological evidence for the pathogenesis of MD and to explore the potential role of middle ear dysbiosis in inner ear disorders. Methods Population Ten patients of definite MD who received endolymphatic sac surgery at the Department of Otolaryngology‐Head and Neck Surgery Peking University People's Hospital were included as ME group. The diagnosis of MD was established according to the American Academy of Otolaryngology‐Head and Neck Surgery (AAO‐HNS) classification updated by the Barany Society's Classification Committee in 2015 [14]. And 10 patients diagnosed with CSOM according to Chinese clinical and surgical classification guidelines for chronic otitis media [15] were enrolled as the positive control group (OE). The exclusive criteria included (1) use of systemic or topical antibiotics within 2 weeks before surgery; (2) with immunodeficiency or current use of immunosuppressive therapy; (3) pregnancy; (4) for MD patients, the presence of other known causes of vertigo (e.g., large vestibular aqueduct syndrome, acoustic neuroma); (4) for CSOM patients, histopathological diagnosis of cholesteatoma. This study was approved by the Ethics Committee of Peking University People's Hospital (No. 2025PHB419). All participants provided written informed consent prior to their inclusion in the study. All methods were performed in accordance with the relevant guidelines and regulations. Middle ear samples were collected using sterile swabs intraoperatively after opening the mastoid cavity in MD patients or the tympanic sinus in CSOM patients. The contact with the skin of the external auditory canal was carefully avoided. The swab tip was placed into a sterile tube, which was immediately stored in liquid nitrogen until DNA extraction. 16S rRNA Sequencing and Data Analysis DNA was extracted from samples using a nucleic acid extraction or purification kit (MoBio Laboratories). The V4 region of the bacterial 16S rRNA genes was amplified by PCR using the primers (515F: 5'‐ GTGCCAGCMGCCGCGGTAA‐3', 806R: 5'‐GGACTACHVGGGTWTCTAAT‐3'). Following purification and quantification, the PCR products were subjected to paired‐end (2 × 150 bp) sequencing on the Illumina NovaSeq. 4000 platform. Raw sequencing data from different specimens were identified based on their barcode sequences. Using Vsearch v2.4.4 software, paired‐end reads were merged and filtered. Sequences shorter than 150 bp or with an average quality score below 20 were discarded, and chimeras were removed to obtain effective tags. Effective sequences were clustered into operational taxonomic units (OTUs) at a 97% sequence similarity threshold. Using SILVA 128 as the reference database [16], the taxonomic annotations of the feature sequences were established, and the composition of each sample was determined at each level (phylum, class, order, family, genus, and species). QIIME (v1.8.0) was used to analyze the alpha diversity of each sample within the two groups, including Chao1 index, ACE index, Shannon diversity index, and Simpson diversity index. These parameters majorly reflected the species richness and diversity in the community. R 4.5.1, together with R Studio (version 2025.09.1 + 401) and RTools 4.5, was utilized for analysis. The “vegan” package in R was used for permutational multivariate analysis of variance (PERMANOVA) and the analysis of similarity (ANOSIM). PERMANOVA/ANOSIM analysis based on the Bray–Curtis dissimilarity matrix was used to assess the differences between all samples between the groups. The level of subgroup explanation for sample differences was indicated by the R 2 obtained from the PERMANOVA analysis. For ANOSIM analysis, an R value closer to 1 represented that the differences between groups were greater than those within groups, while a smaller R value indicated that between‐group differences were minor. The permutation multivariate dispersion (PERMDISP) was employed to detect the dispersion of microbial structure of two groups, and p value < 0.05 indicated the difference of heterogeneity was significant between two groups. Linear discriminant analysis (LDA) effect size (LefSe) [17] was utilized to ascertain the impact of species abundance on variation between groups and identify biomarkers that exhibited difference in abundance. The absolute value of LDA > 2 and p value < 0.05 were established as the threshold for significance. Phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt2) was employed to predict and examine the functional differences between the ME and OE groups [18]. Subsequently, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis was used to assess the discrepancies and shifts in the metabolic pathways associated with the functional genes of microbiome in ME and OE [19, 20]. Randonforest analysis was applied to discriminating the samples from different groups using the R package “randomForest” with 1000 trees and all default settings. The generalization error was estimated using 10‐fold cross‐validation. The Bugbase tool was employed for measuring functional phenotypes in the microbiome [21]. Results Demographic and Clinical Statistics Ten patients of the ME group and 10 patients of the OE group were included. Two groups showed no significant difference in the mean age (ME: 49.1 ± 3.86 years vs. OE: 45.7 ± 4.98 years, p > 0.05) or gender distribution. (ME: 40% female vs. OE: 80% female, p > 0.05). Bacteria isolated from the ear smears of OE patients included Methicillin‐sensitive Staphylococcus aureus (MSSA) (2 cases), Methicillin‐resistant Staphylococcus epidermidis (MRSE) (1 case), Klebsiella oxytoca (1 case), Klebsiella pneumoniae (1 case), Serratia marcescens (1 case), and Corynebacterium striatum (1 case). Sequencing Processing and Quality Control A total of 2,448,679 raw tags were generated by sequencing using the Illumina NovaSeq platform. After filtering and chimera removal via the QIIME v1.8.0, 2,398,437 high‐quality clean tags remained for downstream analysis (Table 1). TABLE 1Summary for sample sequencing data.Sample IDRaw tagsClean tagsEffective rate (%)ME110751210442797.13ME212452912075596.97ME313641213467998.73ME410642010155395.43ME511004910889898.95ME610406410206998.08ME713675013319197.40ME810680810343796.84ME911624911416898.21ME1010597510373997.89OE113442813267798.70OE213906013777899.08OE313976613765598.49OE413725213426397.82OE513772013594798.71OE613607313187896.92OE713996013875999.14OE810943510598796.85OE911160110949498.11OE1010861610708398.59John Wiley & Sons, Ltd. Summary for sample sequencing data. The rarefaction curve analysis for each sample approached a plateau, indicating that the sequencing depth was sufficient for further analysis (Figure 1A). A total of 2098 distinct OTUs were identified across all samples. Venn diagram analysis revealed that 717 OTUs were unique to the ME group, 511 were unique to the OE group, and 870 OTUs were shared between the two groups (Figure 1B). Quality control for sample sequencing data. (A) Rarefaction curves for each sample in the ME and OE groups. (B) Venn diagram of the OTUs from the ME and OE groups. ME, Meniere's disease ear samples; OE, chronic suppurative otitis media ear samples; OUT, operational taxonomic units; OTU, operational taxonomic units. Microbial Diversity Analysis The Shannon index (p = 0.0029) and the Simpson index (p = 0.0068) exhibited significant differences in alpha diversity, while the Chao1 index (p > 0.05) showed no significant disparity between two groups (Figure 2). These results generally suggested the ME group possessed greater species richness compared to the OE group. Alpha diversity analysis of ME and OE group. Alpha diversity was represented by the Shannon index (A), the Simpson index (B), and the Chao1 index (C). ME, Meniere's disease ear samples; OE, chronic suppurative otitis media ear samples. The PERMANOVA/ANOSIM test performed on the Bray–Curtis dissimilarity matrix exhibited the between‐group differences were significantly higher than the within‐group differences (R 2 = 0.107, p = 0.002; R = 0.144, p = 0.018) (Figure 3A). However, the visualization via principal coordinates analysis (PCoA) and non‐metric multidimensional scaling (NMDS) showed that the confidence ellipse of OE microbiome almost covered that of ME (Figure 3B,C). Further analysis using the PERMDISP procedure exhibited the ME group exhibited significantly lower heterogeneity than the OE group (p = 0.002) (Figure 3D). Beta analysis of the ME and OE group. The ANOSIM (A), the PCoA (B) and the NMDS analysis (C) were performed based on the Bray–Curtis distance. (D) The PERMDISP analysis. ME, Meniere's disease ear samples; OE, chronic suppurative otitis media ear samples. Taxonomic Profiling and Identification of Specific Bacteria At the phylum level, both the ME and OE groups showed the highest abundance of all bacteria in Firmicutes, with relative abundances of 38.2% and 37.8%, respectively (Figure 4A). At the class level, the highest abundance was Clostridia (32.2%) in the ME group and Gammaproteobacteria (24.6%) in the OE group (Figure 4B). At the order level, the order with the highest abundance in the ME and OE groups was Clostridiales (32.2% vs. 20.3%) (Figure 4C). At the family level, the most abundant taxon in the ME group was Bacteroidaceae (14.7%), while in the OE group, it was Enterobacteriaceae (11.7%) (Figure 4D). The ME group exhibited a high relative abundance of Bacteroides (14.7%) at the genus level, whereas the OE group showed that Bacteroides and Corynebacterium were the most abundant (5.7%) (Figure 4E). The bacterial compositions of the ME and OE groups. (A) Phylum. (B) Class. (C) Order. (D) Family. (E) Genus. ME, Meniere's disease ear samples; OE, chronic suppurative otitis media ear samples. At the genus level, Ruminococcus, Ochrobactrum, and Halomonas were enriched in the MD group, and the difference reached marginal significance after adjustment (Table 2). TABLE 2Bacteria with abundance differences between two groups.TaxonomyME (%)OE (%) p valueAdjusted p value Ruminococcus 1.250.310.0040.063 Ochrobactrum 1.010.130.0010.052 Halomonas 5.010.990.0020.052John Wiley & Sons, Ltd. Bacteria with abundance differences between two groups. The LEfSe analysis identified Bacteroides, Halomonas, and Prevotella at the genus level, which were significantly enriched in ME, while Neisseriales at the order level and its corresponding family Neisseriaceae were downregulated. (|LDA> 2|, p < 0.05) (Figure 5) LEfSe analysis of ME and OE group (|LDA score| > 2, p < 0.05). (A) The bar chart showed abundant bacteria in ME and OE group. (B) The branching evolution chart exhibited the significantly enriched taxa at different levels. ME, Meniere's disease ear samples; OE, chronic suppurative otitis media ear samples. The Random Forest analysis further identified Ochrobactrum at the genus level and its corresponding family Brucellaceae, along with Hyphomicrobiaceae at the family level and the corresponding order Rhizobiales, and Bacteroidaceae at the family level as key features for classifying the samples (Figure 6). The Random forest analysis of ME and OE groups at all levels. ME, Meniere's disease ear samples; OE, chronic suppurative otitis media ear samples. Predicted Functional Profile and Metabolic Pathway Analysis Microbial phenotypes predicted using BugBase revealed significant functional differences between the ME and OE microbiomes (Table 3). The proportion of facultatively anaerobic bacteria was significantly lower in the ME group (6.73%) compared to the OE group (23.62%, p = 0.019). While not reaching statistical significance, the proportion of anaerobes was higher in the ME group (65.25% vs. 45.30% in the OE group, p = 0.123). Furthermore, the predicted prevalence of bacteria containing mobile genetic elements, which are often associated with virulence and antibiotic resistance, was significantly lower in the ME group (16.80%) compared to the OE group (31.15%, p = 0.029). There were no significant differences in the predicted proportions of potentially pathogenic bacteria or those capable of forming biofilms. TABLE 3Predicted microbial phenotypes based on BugBase analysis.Predicted phenotypeME (%)OE (%) p valueAerobic23.0124.620.579Anaerobic65.2545.300.123Facultatively anaerobic6.7323.620.019* Contains mobile elements16.8031.150.029* Forms biofilms26.3037.080.218Potentially pathogenic11.5421.180.739Stress tolerant13.0422.680.579* p < 0.05.John Wiley & Sons, Ltd. Predicted microbial phenotypes based on BugBase analysis. p < 0.05. The abundance of the metabolic pathways at level 3 based on the KEGG database was analyzed to compare the differences in metabolism between the two groups. Two hundred twenty‐five metabolic pathways were annotated, and 9 metabolic pathways were significantly different between the two groups. However, the disparity did not reach statistical significance after adjustment (Table 4). TABLE 4Analysis of different KEGG metabolic pathway enrichment.KEGG metabolic pathwaysME (%)OE (%) p valueAdjusted p valueMetabolism of xenobiotics by cytochrome P4507.142.860.020.58Drug metabolism ‐ cytochrome P4506.903.100.020.58Flavone and flavonol biosynthesis6.273.730.020.58Glycosphingolipid biosynthesis ‐ globo series8.281.720.020.58Isoflavonoid biosynthesis5.844.160.000.24Linoleic acid metabolism6.163.840.020.58Other glycan degradation6.263.740.000.24Primary bile acid biosynthesis6.433.570.000.24Secondary bile acid biosynthesis6.283.720.020.58John Wiley & Sons, Ltd. Analysis of different KEGG metabolic pathway enrichment. Discussion This study systematically analyzed the MEM in patients with MD, utilizing CSOM as a positive control. Results indicated that MD patients exhibited microbial dysbiosis in the MEM, characterized by significantly higher alpha diversity and lower heterogeneity compared to CSOM patients and an enrichment of low‐virulence, opportunistic, and anaerobic bacteria. This discovery provided novel microbiological evidence for the pathogenesis of MD, suggesting that inner ear inflammation driven by middle ear dysbiosis might play a crucial role in the pathophysiological process of MD. This study found that the alpha diversity of MEM in MD was significantly higher than that of the CSOM group. In diseases driven by infection, the proliferation of dominant pathogens typically suppresses other microorganisms, leading to decreased alpha diversity. This was observed in the MEM of cholesteatoma [12] and otitis media with effusion (OME) [22]. In contrast, in patients with MD, the coexistence of multiple low‐virulence, opportunistic, and immunomodulatory bacteria might exert an inflammatory stimulus on the middle ear and inner ear, which was consistent with the chronic, relapsing course of MD. While ANOSIM/PERMANOVA indicated a significant difference in microbial structures between two groups, the small effect sizes (R = 0.144, R 2 = 0.107) and the PCoA/NMDS plots, where OE samples almost enveloped the ME samples, suggested the compositional separation was minimal. This indicated that MD and CSOM might share a similar microbial spectrum rather than being totally different in bacterial communities. The following PERMDISP analysis (p < 0.01) revealed that the disparity was primarily driven by unequal intra‐group dispersion, demonstrating that the intra‐group variability was significantly lower in ME group. This suggested the MD microbial profile was more stable and homogenous compared to OE group. The analysis of taxonomic composition exhibited that at the phylum level, two groups were enriched by Firmicutes, consistent with the healthy MEM [12]. At the genus level, ME were enriched by opportunistic pathogens including Bacteroides (14.7%), Halomonas (5.0%), Prevotella (4.5%), and Parabacteroides (4.0%). Bacteroides and Parabacteroides were known for immunomodulatory and pro‐inflammatory effects [23, 24], and Bacteroides was detected in over 10% of otitis media sample cultures [25]. Bacteroides fragilis could induce abscess and tissue destruction [23], and Parabacteroides distasonis was observed in extra‐intestinal abscess formation [26, 27, 28]. Halomonas, a genus of Gram‐negative halophilic bacteria, was reported at an abundance of 7.85% in the middle ear effusions of OME children [29]. Given the halophilic nature of Halomonas, the impaired sodium‐regulated ion transport and resulting osmotic shifts associated with endolymphatic hydrops in Ménière's disease might potentially create a favorable hyperosmotic microenvironment for its colonization [30]. Prevotella, one of the core anaerobic genera of the oral cavity [31], can activate Th17‐mediated inflammation via Toll‐like receptor 2 (TLR2) and is involved in inflammatory diseases such as periodontitis and ulcerative colitis [32]. Further analysis identified Ruminococcus, Ochrobactrum, and Halomonas as potential biomarkers for the MD patients. Ruminococcus is a genus of Gram‐positive, strictly anaerobic gut commensals that triggered inflammation cascades via Toll‐like receptor 4 (TLR4) activation and became dominant in the gut microbiome during acute exacerbations of inflammatory bowel disease [33]. Ochrobactrum is a genus of aerobic Gram‐negative rod‐shaped bacteria that served as an opportunistic pathogen [34] occasionally caused severe infections such as endocarditis and bacteremia [35]. We postulated that the MD microbial profile characterized by opportunistic pathogens resulted from previous or latent otitis media. Past infections likely caused transient Eustachian tube dysfunction and impaired middle ear mucosa, leading to the colonization of these specific anaerobic, biofilm‐forming bacteria, which translocated from the nasopharynx [36]. BugBase phenotypic analysis suggested the proportion of facultative anaerobes, including common otitis media pathogens (e.g., Streptococcus pneumoniae, Haemophilus influenzae, Staphylococcus aureus) [13], was significantly lower in the ME than in the OE (6.73% vs. 23.62%, p = 0.019). The relative lack of mobile genetic elements in the ME microbiome (16.80% vs. 31.15%, p = 0.029) indicated the evolutionary environment was less challenging compared to OE group [37]. These findings further supported that the MEM of MD patients was dominated by low‐virulence opportunistic pathogens, suggesting a unique and stable microbiological structure distinct from the acute infective profile of CSOM. Although not statistically significant, the higher abundance of anaerobes in the MD group (65.25% vs. 45.30%, p = 0.123) further indicated the possibility of middle ear hypoxia. In 1997, R. S. Kimura et al. found that improving middle ear ventilation could significantly delay the development of endolymphatic hydrops in guinea pigs and postulated that ameliorated middle and inner ear oxygenation and the release of middle ear pressure reduced endolymphatic pressure [38]. Furthermore, Thomsen J et al. found that modified radical mastoidectomy (MRM), as a placebo surgery, had a control rate for vertigo comparable to that of endolymphatic sac decompression in MD patients [39]. We speculated that the effect of MRM for MD might be attributed to its facilitation for middle ear ventilation. A recent report exhibited that enlarging the tympanic antrum and clearance of the mastoid air cells achieved a favorable vertigo control rate in MD patients [11], supporting the possibility of middle ear drainage obstruction. In contrast, the OE group included in this study presented with tympanic membrane perforation, which objectively improved ventilation of the tympanic cavity and suppressed the anaerobes. We hypothesize that middle ear hypoxia caused by ventilation impairment may select for an adaptive anaerobic microbiome, whose metabolic activities could exacerbate local hypoxia and inflammation, thereby participating in MD pathogenesis. Our research suggested the inner ear pathology of MD might be attributed to the microbial dysbiosis in the middle ear, challenging the conventional view of MD as a sterile inner ear inflammation. Previous studies exhibited that intratympanic LPS injection induced otitis media and increased the permeability of the blood‐labyrinth barrier (BLB), progressively leading to hair cell loss, stria vascularis atrophy, and spiral ganglion damage [40, 41, 42]. Based on these findings, we proposed a “Middle Ear‐Inner Ear Inflammatory Axis” model for MD. We hypothesized that bacterial products from the abnormal MEM led to persistent release of pro‐inflammatory cytokines such as TNF‐α. These products may then permeate the round window membrane into the inner ear, impairing BLB function, exacerbating local microcirculatory disturbances and hypoxia, and ultimately causing endolymphatic hydrops and damage to hair cells and spiral neurons, thereby resulting in MD. In conclusion, the microbiome of ME middle ear exhibited higher community richness and was relatively homogenous in different samples. The difference in bacterial profile between two groups majorly resulted from the intra‐group variety of OE group, and the microbial composition of MD was actually similar to that of CSOM. The ME middle ear microenvironment was characterized by the enrichment of several anaerobic opportunistic pathogens. These results provided novel perspectives into the MEM and mechanisms of MD, challenging the conventional concept of MD as a sterile inner ear disease and suggesting inner ear inflammation ignited by middle ear dysbiosis may contribute to its pathogenesis. However, the small sample size restricted the statistical power, and the insignificantly enriched metabolic pathways involving immunomodulatory effects, hair cell apoptosis [43, 44], and metabolic homeostasis [45] should be further explored and validated. Metagenomic sequencing study with a large sample size and a healthy middle ear control group is needed for further investigation concerning the impact of microbiome on MD pathogenesis. Author Contributions Hong‐Wei Zheng and Li‐Sheng Yu: conceptualization. Lin Wang and Yu‐Ting Tao: data curation. Yu‐Ting Tao: formal analysis. Hong‐Wei Zheng and Li‐Sheng Yu: funding acquisition. Lin Wang: methodology. Hong‐Wei Zheng: project administration. Yu‐Ting Tao: visualization. Yu‐Ting Tao: writing – original draft. Lin Wang, Li‐Sheng Yu, and Hong‐Wei Zheng: writing – review and editing. Ethics Statement This study was approved by the Ethics Committee of Peking University People's Hospital (No. 2025PHB419). All participants provided written informed consent prior to their inclusion in the study. All methods were performed in accordance with the Declaration of Helsinki. Conflicts of Interest Professor Li‐Sheng Yu is a member of World Journal of Otorhinolaryngology ‐ Head and Neck Surgery (WJOHNS) editorial board and is not involved in the peer review process of this article. The other authors declare no conflicts of interest. Declaration of Generative Artificial Intelligence (AI) Tools in the Writing Process During the generation and presentation of this work, the authors used Gemini 2.5 pro version to improve the language. The authors reviewed and refined the content and claimed to be accountable for the originality, validity, and integrity of submissions.