Genomics of Otitis Media (OM): Molecular Genetics Approaches to Characterize Disease Pathophysiology

Otitis media (OM) is an infective and inflammatory disorder known to be a major cause of hearing impairment across all age groups. Both acute and chronic OM result in substantial healthcare utilization related to antibiotic prescription and surgical procedures necessary for its management. Although several studies provided evidence of genetics playing a significant role in the susceptibility to OM, we had limited knowledge about the genes associated with OM until recently. Here we have summarized the known genetic factors that confer susceptibility to various forms of OM in mice and in humans and their genetic load, along with associated cellular signaling pathways. Spotlighted in this review are fucosyltransferase (FUT) enzymes, which have been implicated in the pathogenesis of OM. A comprehensive understanding of the functions of OM-associated genes may provide potential opportunities for its diagnosis and treatment.


INTRODUCTION
Otitis media (OM) is defined as an infective and inflammatory disorder of the middle ear. While OM is associated with significant heterogeneity in clinical presentation, the broad types include acute otitis media (AOM), chronic suppurative otitis media (CSOM), and chronic otitis media with effusion (OME). Previous studies have shown that the pooled incidence of AOM is about 11% worldwide, with 51% of cases occurring in children under the age of five (Monasta et al., 2012). The recurrence of AOM may lead to OME, which has a worldwide incidence rate of up to 5% (Monasta et al., 2012). Both AOM and OM continue to be associated with healthcare utilization in the form of antibiotic therapy, physician and emergency room visits, and common surgical procedures such as tympanostomy tubes, although the advent of effective antimicrobial therapy has led to a substantial reduction in the burden of CSOM (Thomas et al., 2004). A smaller number of studies have also described an association between middle ear infections and speech and language deficits, emphasizing the role of OM in childhood development (Roberts et al., 2004).
The most common bacteria isolated from the middle ear of patients with AOM include Streptococcus pneumoniae and Haemophilus influenzae, although Moraxella catarrhalis, Streptococcus pyogenes, and Staphylococcus aureus are less frequently observed. In contrast, Pseudomonas aeruginosa and S. aureus are the most frequently observed pathogens in CSOM (Giebink and Canafax, 1991).
OM is a multifactorial disorder that may be attributed to a combination of etiologic factors including immunologic, genetic, environmental, and anatomic characteristics. Seasonal microbial susceptibility and Eustachian tube dysfunction are the commonly observed causes (Swanson and Hoecker, 1996;Fireman, 1997). It is well-known that viruses from the respiratory airways also play a crucial role in the pathogenesis of AOM (Nokso-Koivisto et al., 2015). Further, exposure to tobacco smoke, the use of a pacifier, and daycare attendance are among the risk factors for OM, while breastfeeding and pneumococcal vaccines have protective effects (Swanson and Hoecker, 1996;Lubianca Neto et al., 2006;Abrahams and Labbok, 2011;Norhayati et al., 2017).
Beyond environmental factors, genetic background also confers susceptibility to OM, although the disease mechanism is not fully understood. Several OM-associated genes, identified through studies in humans and in animal models, are known to play fundamental roles in diverse biological processes including (1) the development of the middle ear cleft and the Eustachian tube, (2) immune response, (3) bacterial adhesion and viral infection rate, (4) regulation of extracellular matrix, and (5) clearance of pathogens (see Tables 1, 2 for specific studies). In this review, we summarized the genomic variants and factors that have been reported in patients with various forms of OM. Early genetic association studies, mouse, mouse-to-man, human candidate, and genome-wide association studies that correlate OM and genetic variations are also briefly discussed. However, we particularly focused on the recent findings of the associations of A2ML1 and FUT enzymes with OM and offered our perspective on the potential disease mechanism that intuitively can lead to OM in individuals harboring variants of FUT2.

Early Studies
The genetic contribution to OM susceptibility became evident in the 1980s after several studies showed that the prevalence of OM was disproportionately high in some ethnicities (native Americans and Australian aborigines) and relatively low in individuals of African ancestry (Clements, 1968;Bhutta, 2015). A surveillance study on ear and nasopharyngeal carriage was conducted among remote Australian aboriginal communities in 2013 and found that 50% of young children (mean age 13 months) had OME, 37% had AOM, and 12% had CSOM (Leach et al., 2016). Today, CSOM continues to be strongly implicated in the prevalence of hearing and learning disorders in Australian aboriginal communities (Morris, 1998).
One of the earliest genetic studies on OM, conducted in 1983, analyzed the blood groups (ABO) in a cohort of 610 children with chronic otitis media with effusion (COME) and concluded that blood group "A" was a genetic risk factor for OM based on their observation of its higher prevalence in children with COME as compared to non-affected children (Mortensen et al., 1983). Later studies have shown that human leukocyte antigen (HLA) 2 and HLA3 are strongly associated with AOM, while patients with COME have a lower frequency of HLA2 (Kalm et al., 1991(Kalm et al., , 1994. The heritability and genetic components of time with and the number of episodes with OME and AOM during the first 2 years of life were also investigated in a twin and a triplet study in 1999 and found a strong association between the duration or the number of episodes of OM and genetic makeup (Casselbrant et al., 1999).
The contribution of genetics to OM susceptibility is supported by studies reporting a higher incidence of OM in children with chromosomal abnormalities. For example, the prevalence of OME in children with Down syndrome approaches 38% (Austeng et al., 2013). Genes present on chromosome 21 in combination with craniofacial defects such as midfacial hypoplasia, short palate, and Eustachian tube dysfunction (Shibahara and Sando, 1989) and defects of the immune system (Ram and Chinen, 2011) observed in children with Down syndrome may contribute to their increased risk of OM. Ets1 gene, encoding a proto-oncogene, has been recently associated with craniofacial abnormalities and OM in a mouse study (see section Mouse and Mouse-to-Man Studies) (Carpinelli et al., 2015). In humans, the ETS2 gene that also belongs to the protooncogene gene family is present on chromosome 21 and may contribute to OM susceptibility in Down syndrome.
Several studies conducted on cohorts with Turner syndrome, a genetic disorder of partial or complete loss of chromosome X in females, described a highly variable (ranging from 9.1 to 91%) incidence of AOM (Sculerati et al., 1990;Bois et al., 2018). While the karyotype analysis did not reveal any significantly high-risk subgroup, females with Turner syndrome also have greater prevalence and longer duration of middle ear pathologies (Gawron et al., 2008;Bois et al., 2018). These findings implicate some of the X chromosome genes in middle ear development, function, or health.

Mouse and Mouse-to-Man Studies
The development and the phenotyping of transgenic and knockout mouse models in the last 30 years have significantly helped to identify several genes and genetic variations that confer susceptibility to OM in mice. Most of these mouse models spontaneously develop OM; studying their ear morphology and function provided insights into the disease pathophysiology at a molecular level. For instance, Eriksson et al. (2006) showed that plasminogen (Plg)-deficient mice spontaneously develop chronic OM by 18 weeks of age. Plasmin, the active serine proteinase enzyme form of PLG, is mainly involved in the dissociation of fibrin clots and promotes the degradation of the extracellular matrix (Ayon-Nunez et al., 2018). Plasmin plays a critical role in several cellular processes, including wound healing, immunity, tissue remodeling, inflammation, and cell migration (Tefs et al., 2006). Recent studies have shown that certain bacteria possess plasminogen-binding adhesions on their cell surface to exploit the fibrinolytic system, facilitating bacterial uptake and invasion (Raymond and Djordjevic, 2015;Ayon-Nunez et al., 2018).
The role of transcription factors in OM pathology became apparent through the studies of mutant mice lacking Eya4, Evi1, Tgif, Ets1, and Fli1 genes Parkinson et al., 2006;Depreux et al., 2008;Tateossian et al., 2013;Carpinelli et al., 2015). Mice lacking Eya4 have Eustachian tube dysfunction, leading to an increased incidence of OME and hearing impairment (Depreux et al., 2008). Variants in Evi1 in Junbo mice have been shown to cause susceptibility to CSOM. Junbo mice accumulate middle ear effusions and develop hypoxia, inflammation, and thickening of the mucoperiosteum (Parkinson et al., 2006;Bhutta et al., 2014). Later studies have shown that the loss of BPIFA1, one of the most abundant secretory proteins in the upper respiratory tract (Musa et al., 2012), exacerbates the severity of OM in Junbo mice. While Bpifa1 mutant mice did not show any OM susceptibility, the deletion of Bpifa1 in mice carrying Evi1 Junbo variant leads to the thickening of the middle ear mucosa and an increase of collagen deposition (Mulay et al., 2018). Loss of Tgif1, which encodes for TGIF1, results in OME accompanied by the thickening of the middle ear epithelial lining, an increase of goblet cell population, elevated levels of TNFα and IL-1β in ear fluids, and conductive hearing loss in mice (Tateossian et al., 2013). Similarly, haploinsufficiency for Ets1 and Fli1 in mice results in craniofacial abnormalities, including a smaller middle ear cavity and fusion of ossicles to the walls of the middle ear (Carpinelli et al., 2015). Furthermore, Fli1 ± and Ets1 ± double-mutant mice have hearing impairment and their middle ear mucosa is infiltrated by proinflammatory cells, leading to OM (Carpinelli et al., 2015). Hardisty et al. (2003) showed that Jeff mutant mice carrying a Fbxo11 variant have craniofacial abnormalities, elevated hearing thresholds, and middle ear effusion. Defects in the bulla cavitation were observed in Fbxo11 mutant mice, which ultimately result in middle ear adhesions and soft tissue mineralization of the bony anatomy (Del-Pozo et al., 2019). Using N-ethyl-N-nitrosourea mutagenesis, Crompton et al. (2017) showed that the pathogenic variant, p.Leu972Pro, also known as edison variant, in the Nischarin (Nisch) gene leads to mild craniofacial defects, spontaneous OM by 20 weeks, and progressive hearing loss. Recent studies have reported the association of TGIF1 and NISCH loci as potential risk areas for OM in humans (Bhutta et al., 2017), thus supporting the relevance of knowledge obtained from mouse models to the pathophysiology of OM in humans.
Finally, Eda and Edar transcription factors mutant mice (Eda Ta and Edar dlJ/dlJ ) also developed chronic rhinitis and OM (Azar et al., 2016). In these mutants, the nasopharyngeal glandular epithelium fails to develop, which leads to the loss of lysozyme secretion, the reduction of mucociliary clearance, and the overgrowth of commensal bacteria. The spread of nasal S. aureus in Eda Ta mice and of Escherichia coli in Edar dlJ/dlJ mice into the middle ear bulla potentially triggers inflammation and OM (Azar et al., 2016). A non-exhaustive list of the most common loci associated with OM in mouse is presented in Table 1.

Human Candidate Gene-Based and Genome-Wide Association Studies
While early candidate gene-based OM studies have been done mostly on Caucasian patients with recurrent AOM and chronic OME (see section Early Studies), more recent genetic studies have been focusing on ethnic groups or communities for which marriages within the families are

Candidate Gene-Based Studies
Many variants in the genome have been associated with infectious diseases (Klebanov, 2018). In some instances, the clinical features and the biological mechanisms -such as immune response, inflammation, bacterial adhesion, viral infection, and mucociliary clearance -involved in those infectious diseases are compatible with the mechanisms involved during an episode of OM, thus marking the genes known for these disorders as prime candidates for OM susceptibility and recurrence. Candidate gene-based studies on OM have mainly involved genes associated with innate immunity and inflammation (Sale et al., 2011). Those studies have been performed on cohorts from all over the world including US, Finland, Australia, Netherlands, Greece, and Belgium and have identified over 100 alleles that confer susceptibility to various forms of OM (see Table 2 for the partial list of these alleles, genes, and associated clinical features).

A2ML1
In a large consanguineous indigenous Filipino pedigree with a high frequency of OM, Santos-Cortez et al. (2015) showed, by whole-exome and Sanger sequencing, that an 8 bp duplication in the A2ML1 gene (LOD score of 7.5) was associated with susceptibility to OM. The same duplication was found in a heterozygous or a homozygous fashion in three individuals (European American and Hispanic) of another cohort of 123 children prone to OM and absent in non-otitis-prone children and more than 62,000 next-generation sequences. The authors identified seven additional heterozygous A2ML1 variants in patients of European American and Hispanic American origin with OM. A2ML1 encodes alpha-2-macroglobulin-like protein 1 (A2ML1), a protein that traps proteinases and cleaves them. A follow-up study identified 16 additional A2ML1 variants in OM subjects in indigenous Filipino and Pakistani families and US probands (Larson et al., 2019). Based on the expression in the murine mucosal epithelium of the middle ear, it has been speculated that A2ML1 may have a protective function by regulating the proteases present in the middle ear cavity and may also regulate the desquamation of epidermis (Galliano et al., 2006). Recently, by 16S rRNA sequencing of the microbiota of the middle ear of an indigenous Filipino community prone to OM and segregating A2ML1 variants, although not having a statistically significant difference between the cases and the controls, a taxonomic analysis revealed the relative abundance of the phyla Fusobacteria and Bacteroidetes and the genus Fusobacterium in A2ML1 carriers compared to non-carriers (Santos-Cortez et al., 2016).

FUT2
In the same large consanguineous indigenous Filipino pedigree with high frequency of OM, partly due to variants in A2ML1, a subset of individuals were wild type for A2ML1 but were prone to OM (Santos-Cortez et al., 2015). Further genetic analysis determined that the FUT2 variant (p.Arg202 * , LOD score of 4.0) confer susceptibility to OM in those individuals (Santos-Cortez et al., 2018). Screening of DNA samples from 609 additional multi-ethnic families and simplex case subjects with OM by direct Sanger sequencing, linkage analysis, Fisher exact, and transmission disequilibrium tests revealed several other FUT2 variants (p.Arg138Cys, p.Trp154 * , and p.Ala104Val) that confer susceptibility to OM (Santos-Cortez et al., 2018). FUT enzymes are involved in the protein glycosylation pathway. FUTs transfer an L-fucose sugar derived from GDPfucose (donor substrate) to a protein (acceptor substrate). The FUT family contains 13 members (FUT1-FUT13), and many of the FUTs are essential for the synthesis of blood group antigens. FUTs are single-pass type II membrane proteins, resident to the trans-Golgi, while the catalytic domain of FUT proteins resides in the lumen of the Golgi. In humans, FUT1 and FUT2 encode galactoside 2-L-fucosyltransferase, while FUT3 encodes galactoside 3(4)-L-fucosyltransferase. FUT1 and FUT2 transfer L-fucose onto a β-D-galactosyl-(1→4)-N-acetyl-β-Dglucosaminyl derivative and create the oligosaccharide FuCalpha [(1,2)Gal-beta-], also known as H antigen, which is a soluble precursor essential for the final step in the soluble A and B FIGURE 1 | Role of FUT2 in the regulation of blood group and Lewis antigens at the surface of ciliated cells of the middle ear mucosa: secretor and non-secretor status are illustrated. The imbalance of optimal bacteria (green) and pathogenic bacteria (red) is shown in a non-secretor individual that is prone to acute otitis media. antigen synthesis pathway. FUT3 transfers L-fucose onto a β-Dgalactosyl-(1→4)-N-acetyl-β-D-glucosaminyl derivative, or onto H antigen, in order to create blood group Lewis antigens.
FUT1 and FUT2 are differentially expressed in various cell types. For instance, FUT1 expression is restricted to cells of mesodermal origin (for example, erythrocytes), and FUT2 expression is being restricted to cells of endodermal origin (such as the middle ear mucosal cells). Therefore, A and B antigens will be expressed at the surface of red blood cells under the control of FUT1, while A and B antigens will be expressed at the surface of mucosal cells under the control of FUT2. Genetic variations in FUT1 and FUT2 naturally exist. Some FUT1 and FUT2 variants lead to non-functional enzymes, while certain variations in FUT2 can also lead to a reduction of its expression (Santos-Cortez et al., 2018). For instance, the FUT1 p.Tyr154Cys variant ablates the functional activity of the catalytic domain, resulting in the absence of A, B, or H antigens at the surface of erythrocytes (also known as the Bombay phenotype). Similarly, the FUT2 p.Trp154 * variant causes absence of A, B, or H antigens at the surface of mucosal cells (a.k.a. non-secretor status) (Domino et al., 2001a,b). The p.Trp154 * variant of FUT2 is also FIGURE 2 | Protein pathways contributing to otitis media (OM) in human: hypothetical network of human proteins in the middle ear mucosa that have been associated with OM. When these proteins are dysfunctional due to OM-associated genetic variants, they could potentially have an impact on the microbiota of the middle ear cavity. responsible for the non-secretor phenotype in European and African populations (47 and 42%, respectively).
A, B, and H and Lewis antigens are known to serve as an energy source while also regulating the adhesion of bacteria to the cell surface (Ewald and Sumner, 2018; Figure 1). Intuitively, the different blood group antigens and their quantity at the surface of the cells of the mucosal epithelium of the middle ear would have an impact on the microbiota present in the middle ear cavity, and various blood types have been associated with OM (Wiesen et al., 2019). For instance, studies in a Finnish cohort reported a protective effect of blood type O against recurrent AOM, whereas blood type A was associated with increased risk for chronic OME (Wiesen et al., 2019). When tested in vitro, the four FUT2 variants associated with OM (p.Ala104Val, p.Arg138Cys, p.Trp154 * , and p.Arg202 * ) reduced the A antigen levels, while the two nonsense variants also reduced the FUT2 protein levels. Moreover, Fut2 is transiently upregulated in the murine middle ear after inoculation with the non-typeable H. influenza (Santos-Cortez et al., 2018). It is speculated that the OM-associated FUT2 variants are modifying the middle ear microbiome through the regulation of A antigen levels in the middle ear mucosa, thus conferring susceptibility to OM (Santos-Cortez et al., 2018).
While the microbial richness, structure, and composition differences were not statistically significant between the control individuals and the individuals prone to OM in the indigenous Filipino community segregating FUT2 p.Arg202 * variant, the individuals from a Colorado cohort prone to OM and carrier for the FUT2 p.Trp154 * variant had a relatively high abundance of Lactobacillales and Gamma-proteobacteria in their middle ears (Santos-Cortez et al., 2018). Further studies in animal models are necessary to fully understand the FUT2associated OM mechanism.

CONCLUSION
The current genetic and molecular data revealed the association of OM with deficits in each of the following mechanisms: (1) development of the middle ear cavity and Eustachian tube, (2) immune response, (3) bacterial adhesion and viral infection, (4) regulation of the extracellular matrix, and (5) clearance of the middle ear.
In Figure 2, we have attempted to build a network that encompasses most of the known human proteins that have been associated with OM and our hypothesis about the potential impact on the microbiome of the middle ear cavity when these proteins are dysfunctional due to OM-associated genetic variants. Briefly, ABO and FUT2 are localized in the Golgi apparatus of the cells of the middle ear mucosa; these two proteins together create blood group Lewis antigens, potential sources of energy for microbiome. These antigens also provide an adhesion platform for the microbiota. Similarly, goblet cells secrete MUC5AC, MUC5B, MUC2 SFPTA, SFPTA1, and SFPTD to form mucus and surfactants in the middle ear cavity. Bacteria are present in these secretions and are cleared from the middle ear cavity by ciliated cells in order to maintain a healthy microbiome. Cells from the middle ear also secrete interleukins, chemokines, interferons, and necrosis and growth factors in order to recruit immune cells in the extracellular matrix. These cells fight infection and eliminate dead cells. During the inflammatory stage, the extracellular matrix is remodeled by proteases to allow immune cell infiltration. Those proteases are regulated and inhibited by A2ML1 and PAI1. A pathogenic genetic variation in any of those genes would eventually lead to middle ear infection and OM.
Future studies of the enrichment of certain microbiota in individuals with specific genetic variants may eventually help in identifying patients before chronic OM sets in or in devising a patient-specific treatment paradigm in the future.

AUTHOR CONTRIBUTIONS
AG, SA, AI, IA, SR, and ZA wrote the draft and finalized it.

FUNDING
Research in the Ahmed Laboratory was supported by the National Institutes of Health/NIDCD R01DC012564 and R01DC016295.