To provide an expanded view of the genes and variants associated with MD, we performed a systematic literature review. Facilitated by a web tool, we classified, curated, and annotated most of the genes and PubMed abstracts related to MD. Each abstract was systematically annotated for gene names and symbols. The abstracts were computationally organized by genes to be manually reviewed. After the review, the genes were classified to determine which genes showed evidence of mutations or altered expression, as defined by increased or decreased compared to control patients, in MD. In parallel, a systematic review was also performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. Results of both reviews were combined to create a master list of implicated genes in Meniere's disease. The complete literature search strategy as well as study inclusion and exclusion criteria are documented in a PRISMA flowchart (Supplementary Figure 1).

Inbred CBA/J males and females were purchased from JAX (Stock No. 000656). Breeding pairs were set up to obtain P30 mice for immunohistochemistry and single molecule RNA FISH.

In order to curate a comprehensive list of genes implicated in MD, we utilized parallel methodologies for systematic reviews including the use of a webtool, PubTerm, to perform an unbiased search for implicated genes using the previously noted search terms in PubMed and a manual search of public databases and gray literature using PRISMA guidelines. In total, 389 unique abstracts were identified. Abstracts unrelated to Meniere's disease (n = 196) were excluded resulting in 193 abstracts related to Meniere's disease. Exclusion criteria including non-English language (n = 22), animal studies (n = 21), non-gene outcome defined as an absence of genes studied in relation to Meniere's disease (n = 45), and unrelated to Meniere's disease (n = 28) were applied to a full-text review of these references resulting in 77 references being included for systematic review (Supplementary Figure 1). Full-text reviews of identified reference articles as well as review of attached datasets were utilized to ensure that a comprehensive list of genes was defined from these references. Based on the described search strategy, a total of 832 genes were investigated in relation to MD, and 122 of the genes were reported in more than one study. A table of the identified genes and their references is included (Supplementary Table 1). We have provided descriptive tables summarizing the studies implicating genes in MD (Supplementary Datas 1, 2).

Gene ontology (GO) biological process analysis of these genes implicated in MD identified significant enrichment for cellular metal ion homeostasis (GO:0006875), positive regulation of calcium ion transport into cytosol (GO:0010524), positive regulation of calcium ion transmembrane transport (GO:1904427), equilibrioception (GO:0050957), sensory perception of mechanical stimulus (GO:0050954), and sensory perception of sound (GO:0007605). GO molecular function analysis identified significant enrichment for cytokine activity (GO:0005125), water transmembrane transporter activity (GO:0005372), water channel activity (GO:0015250), chemokine activity (GO:0008009), chemokine receptor binding (GO:0042379), and sodium ion transmembrane transporter activity (GO:0015081). GO cellular component analysis identified significant enrichment for integral component of plasma membrane (GO:0005887), MHC protein complex (GO:0042611), and junctional sarcoplasmic reticulum membrane (GO:0014701). The PANTHER classification identified 20 functional groups and revealed that genes implicated in MD encoded metabolite interconversion enzymes (15.1%, including Mif and Sod2), transporters (14.3% including Atp1b2 and Trpv4), intercellular signaling proteins (12.6%, including Tgfb2), protein modification proteins (10.5%, including Wnk2, Wnk4, and Sgk1), transmembrane signaling receptors (10.1%, including Adrb2), and others (Figure 2).

Heatmaps demonstrating genes reported in MD with cell type specificity in both the single-cell and single-nucleus RNA-sequencing datasets are shown in Figures 3A,B (single-cell RNA-Seq) and Figures 4A,B (single-nucleus RNA-Seq), respectively. In the single-cell RNA-seq dataset, fibrocytes were detected and the number of spindle and root cells captured did not enable their transcriptional profiles to be distinguished. In the single-nucleus RNA-seq dataset, spindle and root cell transcriptional profiles were distinguishable, Reissner's membrane cells were detected, and fibrocytes were not detected. Stria vascularis from adult mice were collected, attempting to remove as much of the spiral ligament from the strial tissue. Fibrocytes that were detected represent contaminating cells from the spiral ligament. In both datasets, marginal cells, intermediate cells, basal cells, macrophages, B cells and neutrophils were detected. Details related to these datasets have been described previously (7). Heatmaps demonstrating expression of genes without cell type-specific expression are shown in the supplement for single-cell RNA-seq (Supplementary Figures 2–4) and single-nucleus RNA-Seq (Supplementary Figures 5–9), respectively.

Genes with preferential expression in SV marginal cells include Kcne1, Atp1b2, Esrrb, Add2, Sgk1, Atp13a5, Hmx2, Cacna2d1, Car12, Eya4, Tnfrsf12a, Shroom3, Wnk2, and Dtna (Figures 3A, 4A). Dot plots demonstrate differential expression amongst the major strial cell types including marginal, intermediate and basal cells, as well as some rarer cell types, including SV spindle and root cells, and SV macrophages (Figures 5A,B). In each dot plot, the more orange the dot, the greater the expression level of a given gene and the larger the dot, the greater the proportion of cells that express the given gene. Genes with previously published expression in marginal cells, include Kcne1 (2, 43), Atp1b2 (2, 44–47), Esrrb (7, 48), and Atp13a5 (7). Examination of SGK1 protein expression in the rat and guinea pig cochlea demonstrates expression in the stria vascularis in addition to the spiral ganglion neurons, spiral limbus, organ of corti, and spiral ligament (49, 50). Examination of in situ hybridization of Sgk1 in the E15.5 mouse in the Allen Brain Atlas suggests localization to the organ of Corti and the roof of the cochlear duct where future marginal cells reside (Supplementary Figure 10A). Hmx2 has been previously localized to the developing mouse stria vascularis (51). Cacna2d1 has been previously localized to spiral ganglion neurons (52) and examination of in situ hybridization in the E15.5 mouse in the Allen Brain Atlas suggests possible localization to the roof of the cochlear duct where future marginal cells reside (Supplementary Figure 10B). Carbonic anhydrases have been previously shown to be expressed in the stria vascularis (53–55) and Car12 expression has been previously shown to be expressed in root cells with reduced expression in marginal cells (56). Eya4 has not been previously localized to the mouse stria vascularis but examination of in situ hybridization in the E15.5 mouse in the Allen Brain Atlas suggests widespread expression in the cochlear duct including the region of the future stria vascularis (Supplementary Figure 10C). Add2, Tnfrsf12a, Shroom3, Wnk2, and Dtna have not been previously localized to structures in the cochlea.

Genes with preferential expression in SV intermediate cells include Met, Cdc14a, Ednrb, Tmem176a, Tmem176b, and Gsta1 (Figures 3A, 4A). Dot plots demonstrate differential expression amongst marginal, intermediate, and basal cells in the SV (Figures 5A,B). Met has been previously shown to be expressed in intermediate cells in the developing and adult mouse SV (7, 57). Cdc14a has been previously shown to be expressed in hair cell stereocilia and hair cells, supporting cells, the osseous spiral lamina, and spiral ganglion neurons (58). Ednrb has been previously localized to the SV but the specific cell type expressing Ednrb could not be defined (59). Tmem176a, Tmem176b, and Gsta1 have not been previously localized to structures in the cochlea.

Genes with preferential expression in basal cells within the SV include Wnk4, Col11a2, and Slc44a2 (Figures 3B, 4B). Dot plots demonstrate differential expression amongst marginal, intermediate, and basal cells in the SV (Figures 5A,B). These genes are co-expressed by SV basal cells which are identified by Cldn11 expression that has been previously demonstrated in both adult mouse and humans (7, 60, 61). Col11a2 has been previously shown to be expressed in the spiral ligament (SL) and has been associated with a variety of syndromic as well as non-syndromic sensorineural hearing loss including both autosomal-dominant (DFNA13) and autosomal-recessive (DFNB53) forms (62–64). Slc44a2, formerly known as choline transporter-like protein 2 (CTL-2), is a multi-transmembrane protein originally discovered as a target of antibody-induced hearing loss (65) and more recently implicated in Meniere's disease (28). Others have suggested that because of its prominent expression in cells facing the scala media, SLC44A2 may play a role in cochlear ionic homeostasis (65). Wnk4 is a known regulator of claudins and paracellular chloride ion permeability (66–68). Wnk4 has not been previously localized to SV basal cells. In summary, these data demonstrate RNA expression of Meniere's candidate genes in adult mouse SV cell types.

To localize genes implicated in Meniere's disease to the stria vascularis, we performed smFISH. Having previously validated expression of Esrrb RNA in adult SV marginal cells (7), we demonstrate co-expression of this Meniere's disease-implicated gene with Atp1b2 (Figures 6A,A′) and Kcne1 (Figures 6B,B′) in adult mouse SV marginal cells. The RNA of both Atp1b2 (Figure 6A) and Kcne1 (Figure 6B) localizes predominantly to the marginal cell nuclei. Images without DAPI labeling are shown to emphasize the co-localization of Esrrb with Atp1b2 (Figure 6A′) and Kcne1 (Figure 6B′), respectively. Missense mutations in ESRRB have been identified in patients with Meniere's disease (27) as well as an autosomal recessive non-syndromic sensorineural hearing loss DFNB35 (48). ATP1B2 is upregulated in peripheral blood mononuclear cells (PBMCs) obtained from patients with Meniere's disease compared to patients without Meniere's disease (23). Single nucleotide polymorphisms (SNPs) in KCNE1 have been identified in patients with Meniere's disease (29, 69–72). However, Campbell and colleagues, in comparing two larger cohorts of Meniere's disease and control patients in the Caucasian population, failed to find a significant association between several KCNE1 SNPs and Meniere's disease (73). In intermediate cells, we co-localize expression of Met, Ednrb, and Tmem176a RNA (Figures 6C–E). Met (in turquoise), Ednrb (in red), and Tmem176a (in green) RNA are localized to the intermediate cell layer of the stria vascularis (Figure 6C). Image without DAPI labeling (Figure 6C′) is shown to emphasize co-expression of Ednrb, Tmem176a, and Met in the intermediate cell layer. Yellow dashed line boxes delineate two representative regions (region #1 and region #2) that are enlarged to serve as representative examples of Ednrb, Tmem176a, and Met RNA co-localizing to intermediate cell nuclei within the SV (Figures 6D,E, respectively). While SNPs of uncertain significance for MET were identified in patients with familial Meniere's disease (22), PBMCs from Meniere's patients demonstrate increased expression of both EDNRB and TMEM176A compared to control patients (23). These examples validate the ability of these datasets to localize genes implicated in Meniere's disease to specific cell types in the adult stria vascularis.

First, the larger number of genes expressed by marginal and intermediate cells in the SV suggests that these cells may play a more prominent role in MD. The global localization of these implicated genes to major SV cell types is supported by both our previous work in localizing Kcne1, Atp13a5, and Esrrb RNA to marginal cells, Met RNA to intermediate cells, and Slc26a4 and P2rx2 RNA to spindle cells (7) as well as newly localized gene candidates to major adult SV cell types, including Ednrb and Tmem176a (Figure 5). Connexin 26 (GJB2) and Connexin 30 (GJB6) protein have been previously localized to SV intermediate and basal cells in humans (76, 77) and in mouse (78). While localization of expression does not necessarily imply functional significance, the localization of many genes to specific cell types in the SV may suggest that understanding the roles of these cell types may be critical for insight into the underlying pathophysiology of MD.

In localizing Meniere's disease genes to major SV cell types, we suggest that this points to the need to better understand the role that dysfunction in each of these cell types plays in hearing instability and loss. Work by Gallego-Martinez and colleagues suggests the possibility that the accumulation of missense variants in some which include genes expressed by SV cell types may contribute to the development of MD (27). Meniere's disease has an onset of disease in adulthood (79–81) and this suggests a need for inducible mouse models of gene dysfunction which would distinguish the effects of gene dysfunction on hearing in the mature cochlea from those related to dysfunction during development. More generally, mouse models that replicate hearing fluctuation seen in human patients with MD could serve as useful pre-clinical models to examine the mechanism and efficacy of repurposed and novel therapeutics for MD.

Furthermore, investigating dysfunctional calcium homeostasis may be important to understanding mechanisms underlying hearing loss in MD. Gene ontology analysis of genes implicated in MD identified mechanisms involving calcium ion transport as being enriched. Of these genes, single-cell and single-nucleus RNA-Seq datasets demonstrate expression of Atp13a5, Cacna2d1, and Trpv4 in SV marginal cells, Ank2, Cav1, and Wfs1 in SV intermediate cells, and P2rx2 in SV spindle cells. Of these genes, evidence of decreased expression in MD has been noted for Atp13a5, Cacna2d1, and Ank2 (23) while Cav1 expression has been shown to be increased in patients with MD (23, 31). While it has been suggested that TRPV4 expression may be decreased in human endolymphatic sac tissue from MD patients (82), a subsequent case-control replication study examining TRPV4 expression failed to demonstrate an association between TRPV4 expression and MD (83). Expression of TRPV4 in the cochlea in patients with MD has not been compared to unaffected human patients. Missense variants of uncertain significance for WFS1 were identified in patients with MD but a significant excess of these variants was not seen when compared to control populations (27). In contrast, an excess of P2RX2 missense variants was noted in patients with MD (27). In an experimental model for endolymphatic hydrops, Salt and DeMott have previously demonstrated that elevations in calcium concentration in the endolymph, at a time when endolymph volume and the EP are no longer changing, correlates with elevated auditory thresholds (84). These authors suggest that the gradual dysregulation of calcium concentration in the endolymph while associated with endolyphatic hydrops, may be the underlying mechanism of hearing loss in these settings. Providing further evidence to support this theory, Wangemann and colleagues have shown that the loss of Slc26a4 expression in a mouse model for Pendred syndrome results in acidification of the endolymph, a failure of calcium reabsorption from the endolymph, and hearing loss (85). Thus, these data suggest that dysfunctional calcium homeostasis within the endolymph may potentially contribute to mechanisms resulting in the development of MD.

Despite these observations, several caveats apply to the data presented. A large proportion of gene expression changes for genes implicated in MD were determined from expression changes in the peripheral blood mononuclear cells, which may not reflect changes in the inner ear. Furthermore, genomic features, including single nucleotide variants, copy number variants, and other structural variants have not been systematically and uniformly examined in relation to these investigated genes, potentially contributing to bias in the interpretation of the genomic underpinnings of Meniere's disease. Nonetheless, we localize these genes implicated in MD to the SV providing a context for beginning to understand these observed changes and identifying potentially relevant inner ear gene targets. Furthermore, identifiable expression does not necessarily equate to functional importance. Finally, while we acknowledge the limited ability to connect functional attribution of broad expression changes in the blood to mechanisms underlying MD, identifying meaningful gene and cellular targets in the SV establishes a basis for testing hypotheses related to the underlying pathophysiology of MD.

In conclusion, utilizing single-cell and single-nucleus transcriptional profiles, we localize genes implicated in MD to adult stria vascularis cell types. We identify trends in potentially involved SV cell types based on this top-down approach and in doing so, provide justification for the development of inducible cell type-specific models of SV dysfunction as a means of investigating the underlying pathophysiology of MD. Finally, we provide evidence of the reliable ability of our published transcriptional profiles to localize several candidate gene targets to specific SV cell types, establishing a justification for testing the role of these candidate genes in MD.