The genomic landscape of Ménière's disease: a path to endolymphatic hydrops

Abstract

Background: Ménière's disease (MD) is a disorder of the inner ear that causes episodic bouts of severe dizziness, roaring tinnitus, and fluctuating hearing loss. To date, no targeted therapy exists. As such, we have undertaken a large whole genome sequencing study on carefully phenotyped unilateral MD patients with the goal of gene/pathway discovery and a move towards targeted intervention. This study was a retrospective review of patients with a history of Ménière's disease. Genomic DNA, acquired from saliva samples, was purified and subjected to whole genome sequencing.

Results: Stringent variant calling, performed on 511 samples passing quality checks, followed by gene-based filtering by recurrence and proximity in molecular interaction networks, led to 481 high priority MD genes. These high priority genes, including MPHOSPH8, MYO18A, TRIOBP, OTOGL, TNC, and MYO6, were previously implicated in hearing loss, balance, and cochlear function, and were significantly enriched in common variant studies of hearing loss. Validation in an independent MD cohort confirmed 82 recurrent genes. Pathway analysis pointed to cell-cell adhesion, extracellular matrix, and cellular energy maintenance as key mediators of MD. Furthermore, the MD-prioritized genes were highly expressed in human inner ear hair cells and dark/vestibular cells, and were differentially expressed in a mouse model of hearing loss.

Conclusion: By enabling the development of model systems that may lead to targeted therapies and MD screening panels, the genes and variants identified in this study will inform diagnosis and treatment of MD.

Keywords: Gene discovery; Ménière's disease; Network analysis; Systems biology; Whole genome sequencing.

Fig. 1

Summary of variants in study population. A Bar chart displaying variants by type, after application of all filters. B Bar chart displaying variants by function. C Bar chart displaying 35 most frequently mutated genes. D Bar chart displaying the 35 most recurrent variants. E Scatterplot showing the top 50 most frequently mutated genes, with the number of impacted samples on the x-axis, and the ratio of observed frequency to expected frequency on the y-axis. F Scatterplot showing select enriched GO terms in the set of genes with > = 4 variants with obs/exp > 1.3 per gene, ranked by -log(p) (hypergeometric test)

Fig. 2

Network prioritization of candidate genes. A Barplot showing top GO pathways for recurrence + network gene set. B Scatterplot showing the log odds ratio of enrichment between relevant terms in the mammalian phenotype ontology and MD genes filtered in one of three ways: 1) Recurrence = gene has > = 4 unexpected variants (unexpected = obs/exp > 1.3); 2) Network = netprop z > 3; 3) Recurrence + netprop = gene has > = 4 unexpected variants & netprop z > 3 (N = 481 genes). C Barplot showing the highest frequency genes meeting filtering criteria which appear in 2 or more relevant pathways/phenotypes. Right axis (red dotted line) shows cumulative sum of % samples explained by variants in genes. D Subset of the recurrence + netprop set of genes from selected terms and pathways most relevant to MD. Node color indicates the pathway(s) membership. Node size indicates the number of unexpected variants per gene. Medium confidence STRING edges are shown. https://www.ndexbio.org/viewer/networks/c6b7c224-41ed-11ee-aa50-005056ae23aa

Fig. 3

Diagnostic value and validation. A Bar chart showing top replicated genes from external human databases. Right axis (red dotted line) shows cumulative sum of % samples explained by variants in genes. Genes ranked by recurrence. Scatterplot below shows gene set membership. B Enrichment with external human databases: log(obs/exp) + -1SD. Hypergeometric test for enrichment with gene sets shown. *p < 0.05, **p < 0.01, ***p < 0.005. C Cumulative percentage of MD samples explained by UFVs in top 50 recurrent + network (red), ranked by recurrence, compared to the baseline expectation given allele frequencies in general population (black; gnomad). 50% of MD cohort explained by 11 genes. We would expect to recover 33% of a control cohort with these same variants. D Cumulative percentage of MD samples explained by UFVs in high priority genes from pathway and human replication analysis (red), compared to the baseline expectation given allele frequencies in general population (black; gnomad). 36% of MD cohort explained by 33 high priority genes. We would expect to recover 12% of a control cohort with these same variants

Fig. 4

Replication in model of hearing impaired mice A) Number of MD-prioritized genes which are upregulated (red) and downregulated (blue) in ears of hearing impaired mice compared to healthy controls. Data are shown by otic cell types. B Select significantly differentially expressed genes per otic cell type are shown in the heatmap, with red indicating upregulation in the hearing-impaired mice, and blue indicating downregulation in the hearing-impaired mice. HC: Hair Cell, DC_PC: Deiter cell and pillar cell, IPhC_IBC: Inner phalangeal cell/Inner border cell, TBC: Tympanic border cell, Nudt4 + _PC: Nudt4 + pillar cell, EC: Epithelial cell, SGN: Spiral ganglion neuron, SGC: Satellite glial cell, SC: Schwann cell, CC: Chondrocyte, OB: Osteoblast, RMC: Reissner’s membrane cell, PVM_M: Perivascular resident macrophage-like melanocyte, FC1: Fibrocyte1, FC2: Fibrocyte2, FC3: Fibrocyte3, FC4: Fibrocyte4, SMC: Smooth muscle cell, M: Macrophage, T: T cell, B: B cell, Neu: Granulocytes/neutrophils

Fig. 5

Cell-type specific expression of MD genes. A-D UMAP cell types and expression levels of select genes from the human inner ear atlas[36]. POM: periotic mesenchyme. E Relative average expression in human inner ear atlas cell types for select MD prioritized genes which validated in human databases (genes from Fig. 3A). F Violinplots showing the average percentile expression, per human inner ear atlas celltype, for genes which have any variant in the study cohort (baseline; gray), and for genes in the network + recurrent prioritized set (network_recurrent; green). G MERFISH spatial expression of two genes of interest; Otogl, Col11a1, along with Myo7a to indicate hair cells. Panels show i) low magnification view of the cochlea, and zoomed in expression of genes of interest in the utricle (ii) and organ of Corti at the mid-apical turn(OC) (iii). Scale bar: 250 µm. *** FDR < 0.001, ** FDR < 0.01, * FDR < 0.05, ns not significant; wilcoxon rank sum test, benjamini hochberg correction for multiple tests