Hearing Impairment in Mucopolysaccharidosis: A Systems Biology Approach

Mucopolysaccharidoses (MPS) are lysosomal diseases caused by deficiencies in lysosomal enzymes involved in the degradation of glycosaminoglycans (GAGs). Sensorineural hearing impairment is a common feature in MPS patients, but there is no consensus on its etiology. For this reason, we aimed to identify genes and pathways related to hearing loss and to correlate them with gene expression data in MPS. We used HPO and Disgenet to identify candidate genes. We constructed the network with string and Cytoscape, and hub genes were identified in Cytohubba. Expression data were obtained from the MPSBase website. We found the NDUFA gene family as the major hub genes and 114 enriched pathways related to hearing loss. These genes and biological pathways may serve as potential candidates for clinical studies to better understand hearing impairment mechanisms in lysosomal storage diseases like mucopolysaccharidosis.


Gene Candidate Analysis
We searched for the terms "hearing loss" in the Human Phenotype Ontology -HPO [5] and in the Disgenet databases [6]. The HPO database provides a standardized vocabulary of phenotypic abnormalities encountered in human diseases, and their related genes. The Disgenet database has a similar approach, but the database integrates information of human gene-disease association and variant-disease associations from various external databases with information about Mendelian, complex and environmental diseases.
The related terms found in the HPO and Disgenet are shown in Table 1. Then, we used the unique genes between the two databases in the following analysis steps. The flowchart of the methodology is shown in Figure 1.

Network Analysis
A gene network was constructed with the selected candidate genes in the STRING database v.11.0 [7]. We used high confidence network scores (0.07) to obtain only experimental data, and text mining interactions were excluded from the analysis. Only the query proteins were considered, without first and second shell interactors. The analyses were performed in Cytoscape v.3.8, with curated plugins [8]. To identify the hub genes, we used Cytohubba v.0.1 with the local based method Maximal Clique Centrality, MCC [9].

Gene Set Enrichment and Expression Analysis
The functional enrichment was quantitatively assessed (p-value) using a hypergeometric distribution. Multiple test correction was also implemented by applying the FDR at a significance level of p<0.05. We used the Biological Network Gene Ontology (BiNGO) plugin v.3.0.4 [10] to identify the biological processes (BP), molecular function (MF), and cellular component pathways.
To determine the KEGG [11] pathways, we used the pathfindR package [12] in the R environment [13]. To evaluate the expression of the hub genes, we searched for datasets in the MPSBase [14]. We also selected the most frequent genes which appear in the pathways to evaluate their expression in the available transcriptomic MPS datasets, which were obtained from human IPS and Hela cells.

Results
In HPO, we found 1393 genes, and in Disgenet 1078 ( Figure 1). In total, 1679 unique genes were present in either bank. After removing genes without any connections by String, 1617 remained. The Cytoscape network was composed of 827 nodes (genes), and 3777 edges (number of interactions between the genes). The top hub genes and the related neighbors are shown in Figure 2. Most of them are part of the NADH Ubiquinone   Oxidoreductase family, like NDUFB7, NDUFS7, NDUFB8,  NDUFA13, NDUFS2, NDUFV1, NDUFV2, NDUFS3, NDUFA9,  NDUFA2, NDUFB5, NDUFA6, NDUFB9, NDUFB10, NDUFS1, and NDUFA7. In addition, we also identified as hub genes the Mitochondrially Encoded NADH:Ubiquinone Oxidoreductase genes, like MTND6, MTND2, and MTND3. Another hub gene identified in our analysis is the Ubiquinol-Cytochrome C Reductase, Rieske Iron-Sulfur Polypeptide -UQCRFS1 gene.
Regarding the pathway analysis, the most frequent genes were MAPK1, PIK3CA, PIK3R1, AKT1, KRAS, MAP2K1, NRAS, PRKCB, RAF1, and NFKB1. There were 114 enriched pathways related to the hearing loss gene list ( Table 2). The top KEGGrelated pathways are shown in Figure 3. We also constructed the KEGG maps to understand how the gene hub list affects the enriched pathways ( Figure 4, Supplementar File 1).
Gene expression analysis showed NDUF genes (the top hub genes) to be up-regulated in MPS IIIB, while NDFUS7 is downregulated in MPS I. NDUFV2 and NDUFS3 are not identified as differentially expressed. The same pattern is seen in the gene pathways list (Table 3).

Discussion
Several pathways involved in cell adhesion, proliferation and differentiation were enriched in our analysis. The Wnt signaling is the most enriched pathway, as it controls cellular events related to the formation of sensory hair cells during development [15], and in cochlear formation and hair cell differentiation and polarization [16][17]. Other pathways, like PI3K/Akt, MAPK/ ERK and EGFR and ERBB signaling were also enriched. Given that these pathways are related to formation, maintenance and regeneration activity of specialized cells, it is not surprising that these pathways are deranged in progressive degenerative diseases, as MPS. Another set of differentially expressed genes in our analysis were related to mitochondrial function. This organelle has a role in oxidative phosphorylation, oxidative stress control, and apoptosis. The relationship between hearing loss and mitochondrial diseases has been discussed previously in the literature [18][19][20]. Zwirner and Wilichowski demonstrated that there is a high incidence of 42% of sensorineural hearing loss in childrens with mitochondrial encephalomyopathies [21]. Besides, it was shown that causative mitochondrial DNA mutations appear in 5-10% of patients with post-lingual nonsyndromic hearing loss [22].
Mitochondrial defects in MPS were also described in the literature. Martins and collaborators observed structurally abnormal mitochondria and impaired mitochondrial energy metabolism in a 5-month-old mouse model of MPS III C [23]. In another study, light microscopy of brain sections of 6-months-old mice with MPS III B showed the accumulation of mitochondrial ATP synthase subunit c in the brain [24]. Alterations in mitochondria and lysosomes lead to neurological dysfunction and oxidative stress [25], observed in some MPS types. Interestingly, Baixauli et al. [26] showed that mitochondrial deficiency impairs lysosome function, and disrupts endolysosomal trafficking pathways and autophagy, thus linking a primary mitochondrial dysfunction to a lysosomal disturbance. Mitochondrial dysfunction is emerging as a significant contributor to the pathophysiology of lysosomal storage disorders, like MPS [27].
The NADH:ubiquinone oxidoreductase subunits gene family appeared several times in our hub analysis. The Table 3. Gene expression analysis of hub genes and most frequent genes of the pathway analysis. Gene expression data retrieved from https://www.ufrgs.br/mpsbase/ NADH:ubiquinone oxidoreductase (complex I) is part of the respiratory complex and is a major source of reactive oxygen species (ROS) and an essential contributor to cellular oxidative stress. Moreover, ROS production has a relationship with several apoptotic and necrotic cell death pathways in auditory tissues [28]. The subsequent apoptosis induction and elevated ROS formation are involved in developing several hearing loss impairments. In parallel, the involvement of ROS in MPS IVA [29][30] and MPS IIIB [31] has been shown, but the role of the NDUF family in any type of MPS is not yet demonstrated. Several genes related to complex I are up-regulated in MPS III B and down-regulated in MPS I -both of which present hearing loss in a significant portion of patients [32][33][34][35][36]. In MPS III, previous studies demonstrated rates of hearing loss of 87% in MPS III A, 100% in MPS III B, 75% in MPS III C, and 25% in MPS IIID [37][38][39][40]. MPS III B's higher proportion can be related to NDUF expression results, although more studies are needed to understand and validate this relationship.
Other genes found in our network analysis are related to succinate dehydrogenase, fumarate reductase, cytochrome c oxidase and reductase, and V-ATPase, which are all involved in the oxidative phosphorylation ( Figure 4). A3, one of the four isoforms of subunit A V-ATPase, is required for secretory lysosome trafficking to the plasma membrane. It is also necessary to maintain the ionic concentration and pH of the endolymph that bathes the mechanosensory hair cells of the Corti organ in the inner ear [41][42][43][44]. The V-ATPase is also present in the cochlea, and interdental cells are especially V-ATPase-rich [44]. Besides that, mutations in the subunit A4 or B1 are associated with sensorineural hearing loss [43]. Moreover, Santra and Amack, 2021 [45] have shown a specific role for V-ATPase inducing caspase-independent necrosis-like cell death in mechanosensory hair cells in neuromasts. Patients with mutations in specific V-ATPase subunits can develop sensorineural deafness. The mechanism involves modulation of the mitochondrial permeability transition pore, which regulates mitochondrial membrane potential, thus improving hair cell survival.
The majority of young MPS patients present mixed hearing loss (32%) and 16% sensorineural [46]. Many studies exploring the mechanisms of hearing loss in MPS suggest a combination of conductive and sensorineural processes [3]. In this model, GAG accumulation leads to copious secretion and recurrent ear infection that, in conjunction with bone and cartilage deformities, contribute to conductive hearing loss. In addition, sensorineural hearing loss is caused by the death of hair cells. Our study sheds light into this second mechanism, suggesting a role for mitochondrial and V-ATPase dysfunction in the loss of hair cells [47]. Even though we analyzed data from in vitro neural stem cells and not from the inner ear, one can suppose that the same mechanisms that lead to lysosomal storage-derived disturbance of mitochondrial function and V-ATPases in the brain are also present in other cells, such as cochlear hair cells, but in that case with specific consequences. Therefore, experimental studies in MPS animal models could test this hypothesis directly in the involved cells.

Conclusions
We identified several genes and biological pathways involved in ear development and hearing loss. These genes and biological pathways may serve as potential candidates for clinical and experimental studies to better understand hearing impairment mechanisms in lysosomal storage diseases, like MPS.