A Comprehensive Network and Pathway Analysis of Human Deafness Genes

Otology & Neurotology 34:961Y970 Ó 2013, Otology & Neurotology, Inc.

*Georgios A. Stamatiou and †Konstantina M. Stankovic

*Department of Otolaryngology, Hippokration General Hospital, University of Athens, Athens, Greece; and ÞDepartment of Otology and Laryngology, Harvard Medical School and Department of Otolaryngology, Massachusetts Eye and Ear Infirmary, Boston, Massachusetts, U.S.A.

Objective: To perform comprehensive network and pathway factor beta1 (TGFB1) for Group 1, MAPK3/MAPK1 MAP kinase analyses of the genes known to cause genetic hearing loss. (ERK 1/2) and the G protein coupled receptors (GPCR) for Study Design: In silico analysis of deafness genes using inge- Group 2, and TGFB1 and hepatocyte nuclear factor 4 alpha (HNF4A) nuity pathway analysis (IPA). for Group 3. The nodal molecules included not only those known Methods: Genes relevant for hearing and deafness were iden- to be associated with deafness (GPCR), or with predisposition to tified through PubMed literature searches and the Hereditary otosclerosis (TGFB1), but also novel genes that have not been Hearing Loss Homepage. The genes were assembled into 3 groups: described in the cochlea (HNF4A) and signaling kinases (ERK 1/2). 63 genes that cause nonsyndromic deafness, 107 genes that cause Conclusion: A number of molecules that are likely to be key nonsyndromic or syndromic sensorineural deafness, and 112 genes mediators of genetic hearing loss were identified through three associated with otic capsule development and malformations. Each different network and pathway analyses. The molecules included group of genes was analyzed using IPA to discover the most new candidate genes for deafness. Therapies targeting these molecules interconnected, that is, ‘‘nodal’’ molecules, within the most statis- maybeusefultotreatdeafness.Key Words: Deafness genesVERK tically significant networks (p G 10j45). 1/2VGPCRVHNF4AVMAPK1VMolecular pathways analysisV Results: The number of networks that met our criterion for sig- TGFB1. nificance was 1 for Group 1 and 2 for Groups 2 and 3. Nodal molecules of these networks were as follows: transforming growth Otol Neurotol 34:961Y970, 2013.

Hearing loss is the most common sensory deficit in the sensorineural and believed to account for 70% of cases world, affecting almost 600 million people (1), and the (4). The remaining 30% of hearing loss is syndromic most common congenital anomaly, affecting 2 to 6 per and can be sensorineural, conductive, or mixed. More 1,000 newborns (1,2). At least two-thirds of cases of than 400 syndromes include hearing loss as a part of prelingual deafness in the developed countries are due their phenotypic signature (3). Typically, genetic hear- to genetic factors; the remaining one-third is attributed ing loss is monogenic. Based on the mode of inheri- to environmental and unidentified genetic factors (3). tance, monogenic hearing loss is classified as autosomal In general, genetic hearing loss can be classified as recessive, accounting for 80% of cases; autosomal domi- syndromic or nonsyndromic, depending on whether other nant, accounting for almost 20% of cases; X-linked; and distinguishing physical features are present or absent, mitochondrial; the latter two account for less than 1% of respectively. Nonsyndromic hearing loss is typically cases (5). Hearing loss is one of the most genetically hetero- geneous disorders, with more than 100 mapped loci and Address correspondence and reprint requests to Konstantina Stankovic, more than 60 causally implicated genes within these loci M.D., Ph.D., Massachusetts Eye and Ear Infirmary, 243 Charles St, (6). Additional complexity exists because mutations in the Boston, MA 02114; E-mail: [email protected] Study conducted at: Department of Otolaryngology, Massachusetts same gene may cause syndromic or nonsyndromic hearing Eye and Ear Infirmary, Harvard Medical School, Boston, MA, USA. loss, and hearing loss may be oligogenic (7). Given the Source of Funding: NIDCD K08 DC010419 (K. M. S.) and the intricacy of normal hearing, which requires interaction of Bertarelli Foundation (K. M. S.). many diverse molecules, it is estimated that approximately The study was presented at the AOS Spring Meeting, April 21Y22, 2012, San Diego, CA. 1% of human genes play a role in hearing (8). The authors disclose no conflicts of interest. Although the shear number of genes involved in Supplemental digital content is available in the text. hearing and deafness may seem daunting, this complexity

961

Copyright © 2013 Otology & Neurotology, Inc. Unauthorized reproduction of this article is prohibited. 962 G. A. STAMATIOU AND K. M. STANKOVIC may be simplified by looking at well-characterized cell specified molecules. We refer to the most interconnected mol- signaling and metabolic pathways through which these ecule (or class of molecules) in a network as key or central node. genes are known to interact in various tissues. A common Networks of up to 140 molecules were studied to allow for the assumption of pathway analysis is that genes whose dys- possibility that all genes within a group belonged to the same function contributes to a disease phenotype tend to be network. Statistical analysis of networks and pathways was performed as part of the overall IPA using the right-tailed functionally related (9,10). Unraveling these pathways is Fisher’s exact test (11). The p value reflects the likelihood that essential to understanding biological mechanisms, disease the association between the input genes and a given pathway or states, and the function of drugs that affect them. We network is due to chance. Only networks with p G 10j45 were therefore undertook the first comprehensive pathway and considered significant. Less stringent criterion of p G 0.05 was network analysis of all known genes implicated in hu- used for pathway significance to facilitate discovery of man deafness. We used tools of bioinformatics to analyze potentially novel signaling pathways. interactions across multiple biological dimensions includ- Expression of select genes was validated by summarizing ing molecular interactions, cellular processes, and disease their reported expression in the mouse cochlear hair cells based processes. We focused on 3 groups of genes: those causing on RNASeq (12) or mouse spiral ganglion neurons based on nonsyndromic hearing loss, those causing nonsyndromic GeneChips (13). Both databases catalog changes in gene expression during embryonic and early postnatal development. or syndromic sensorineural hearing loss, and those associated with malformation of the otic capsule. The third group of genes was augmented with genes implicated in RESULTS otic capsule development based on animal studies in mammals. Our analyses suggest new candidate genes for Network analysis of the nonsyndromic deafness deafness within the known deafness loci and highlight genes revealed 6 networks. The top network was very several genes as potential novel targets for diagnosis and highly significant (p =10j105) and included nearly treatment of genetic hearing loss. all (50 of 63) genes currently implicated in nonsyndromic HL, suggesting close coupling of deafness genes within the cacophony of human interactome. This network MATERIALS AND METHODS is shown in Figure S1, Supplemental Digital Content 2, A PubMed English search was performed to identify genes http://links.lww.com/MAO/A150, and all molecules of and molecules associated with human hearing loss and with the network are listed in Table S2, Supplemental Digital mammalian development and malformation of the otic capsule. Content 7, http://links.lww.com/MAO/A155. The remain- All genes listed on the Hereditary Hearing Loss Homepage en- ing 5 networks did not meet our criterion for significance tered our analyses. The pertinent genes were identified through (p =10j2). The central node of the top pathway was linkage analysis, immunohistochemistry, quantitative reverse transforming growth factor beta1 (TGFB1), having 35 con- transcription-polymerase chain reaction, or microarray ana- nections with other genes in the network (Fig. 1). The lyses. Only studies with relevant controls or appropriate sta- second and third central nodes were beta-estradiol and tistical analyses were included; studies without control groups or with unspecified statistical significance were excluded. filamentous actin (F-Actin), with 29 and 24 gene inter- Genes that met our criteria were classified based on their estab- connections, respectively (Table 1). Pathway analysis lished role in nonsyndromic or syndromic sensorineural hear- revealed 5 statistically significant pathways (Table 2), ing loss, or association with otic capsule development or highlighting the importance of actin cytoskeleton sig- malformation. The number of genes that entered our anal- naling, which is known to be critical for stereociliar and yses was 63 genes that cause nonsyndromic deafness, 107 genes hair cell function. The second ranked pathway was cell that cause nonsyndromic or syndromic sensorineural deaf- junction signaling (Table 2), consistent with the obser- ness, and 112 genes associated with otic capsule development vations that about 50% of cases of nonsyndromic HL are and malformations (Table S1, Supplemental Digital Content 1, due to mutations in GJB2 encoding a gap junction protein http://links.lww.com/MAO/A149, alphabetically lists genes used (14). The other top pathways point to previously under- in the analyses, along with their associated clinical phenotypes). The second analysis was done to enrich for genes asso- studied pathways in the context of deafness, that is, ciated with sensorineural hearing loss, whereas the third analysis hepatic fibrosis, and signaling via sildenafil (Viagra) or was done to enrich for genes that contribute to mixed conductive integrin-linked kinase (Table 2). and sensorineural hearing loss. Network analysis of genes causing both nonsyndromic For each of the 3 groups of genes, a different analysis was and syndromic sensorineural deafness generated 10 net- performed using ingenuity pathway analysis software (IPA; works, two of which met our criterion for statistical sig- Ingenuity Systems, Redwood City, CA, USA), which is based nificance (with p =10j103 and p =10j46); the other on the world’s largest curated and highest quality knowledge networks had p e 10j2. The central node of the top base of biological networks created from millions of individu- network was MAPK3/MAPK1 MAP kinase (ERK 1/2) ally modeled relationships among proteins, genes, complexes, with 33 connections, closely followed by collagens with cells, tissues, drugs, and diseases to understand how candidate genes may work together as molecular modules that impact 32 connections, and hepatocyte growth factor (HGF) cellular processes and disease phenotypes. Within IPA, pathway with 30 connections (Table 1, Fig. 2, Figure S2, Supple- refers to well-characterized cell signaling and metabolic path- mental Digital Content 3, http://links.lww.com/MAO/A151, ways based on molecular and biochemical studies, whereas showing the top network of the second pathway anal- network refers to regulatory relationships that exist among user- ysis and Table S3, Supplemental Digital Content 8,

Otology & Neurotology, Vol. 34, No. 5, 2013

FIG. 1. A part of the most significant network generated by IPA network analysis of genes implicated in nonsyndromic hearing loss (Analysis 1). The full network is shown in Supplemental Figure 1. Molecules with yellow frames are the top nodes of the network. The input genes are highlighted with blue frames. Molecules within white symbols were provided by the ingenuity knowledge base. Solid and dashed lines between molecules represent direct and indirect interactions, respectively. http://links.lww.com/MAO/A156, listing all molecules http://links.lww.com/MAO/A158, listing all molecules of of the network). The central node of the second most this network). The top 3 nodes of the second most sig- significant network was G Protein Coupled Receptors nificant network were hepatocyte nuclear factor 4 alpha (GPCR) with 74 interconnections (Figure S3, Supple- (HNF4A) with 51 connections, TGFB1 with 48 connec- mental Digital Content 4, http://links.lww.com/MAO/A152, tions, and tumor necrosis factor (TNF) with 31 connec- showing network 2 of the second pathway anal- tions (Table 1, Figure S5, Supplemental Digital Content 6, ysis, and Table S4, Supplemental Digital Content 9, http://links.lww.com/MAO/A154, showing the second net- http://links.lww.com/MAO/A157, listing all molecules work of the third pathway analysis, and Table S6, Supple- of this network). This node was more than 3 times more mental Digital Content 11, http://links.lww.com/MAO/A159, interconnected than the second and third key nodesVbeta listing all molecules of this network). Pathway analysis estradiol with 24 connections and TGFB1 with 21 con- pointed to multiple signaling pathways that may be relevant nections. Pathway analysis pointed to the potential relevance of signaling pathways in melanocytes, melanoma, TABLE 1. The top 3 nodes of the statistically significant and renal cell carcinoma for sensorineural deafness (Table 2), networks identified through 3 different types of analyses in addition to the pathways already highlighted in the first analysis. Network 1 Network 2 Network analysis of genes involved in otic capsule devel- Molecule Molecule opment and malformation syndromes generated 4 net- Analysis (no. connections) (no. connections) works, 2 of which met our criterion for statistical significance 1: Nonsyndromic HL TGFB1 (35) V (with p =10j105 and p =10j91); the other 2 had p =10j2. Estradiol (29) V The central node of the top ranked network was TGFB1 F-actin (24) V 2: Syndromic and ERK 1/2 (33) GPCR (74) with 70 connections, followed by ERK 1/2 and bone nonsyndromic HL Collagen (32) Estradiol (24) morphogenetic protein 4 (BMP4) with 54 and 44 connec- HGF (30) TGFB1 (21) tions, respectively (Table 1, Fig. 3, Figure S4, Supple- 3: Otic capsule development TGFB1 (70) HFN4a (51) mental Digital Content 5, http://links.lww.com/MAO/A153, and malformation ERK 1/2 (54) TGFB1 (48) showing the top network of the third pathway anal- BMP4 (44) TNF (31) ysis, and Table S5, Supplemental Digital Content 10, HL indicates hearing loss.

Otology & Neurotology, Vol. 34, No. 5, 2013

TABLE 2. The top canonical pathways discovered by IPA for each of the 3 analyses Analysis Top canonical pathways p 1: Nonsyndromic HL Actin cytoskeleton signaling 1.22E-04 Tight junction signaling 2.34E-04 Hepatic fibrosis/hepatic stellate cell activation 1.37E-02 Cellular effects of sildenafil (Viagra) 1.47E-02 ILK signaling 2.92E-02 2: Nonsyndromic and syndromic sensorineural HL Melanocyte development and pigmentation signaling 1.19E-05 Actin cytoskeleton signaling 2.86E-04 Melanoma signaling 2.16E-03 Tight junction signaling 2.23E-03 Renal cell carcinoma signaling 8.66E-03 3: Otic capsule development and malformation Role of osteoblasts, osteoclasts and chondrocytes in rheumatoid arthritis 1.4E-15 Factors promoting cardiogenesis in vertebrates 1.24E-10 Human embryonic stem cell pluripotency 6.27E-10 BMP signaling pathway 2.76E-07 Cell cycle control of chromosomal replication 5.22E-07 for deafness and cochlear malformation, including sig- fied as causative (TGFB1, CYP19A1), and their docu- naling pathways in rheumatoid arthritis, cardiogenesis, mented expression in the mouse cochlea. When analyzing stem cells pluripotency, chromosomal replication, and all genes comprehensively across the studied networks BMP signaling (Table 2). (Supplemental Tables S2-S6), many represent additional Chromosomal locations of the top nodes listed in candidate genes for human deafness because of their chro- Table1areprovidedinTable3,alongwiththecorre- mosomal location within deafness loci with unknown sponding human and mouse deafness loci, and the causative genes or within deafness loci where known reported expression in the mouse cochlea. Similar infor- causative genes are different than the candidate genes. mation is provided for all genes in Supplemental Tables Focusing on the former, we have selected 25 genes S2YS6. Of the 10 nodal molecules in Table 3, six repre- whose expression has been unambiguously documented sent novel candidate genes for human deafness because of in murine cochlear hair cells (12) or cochlear neurons (13) their location within human deafness loci with unknown (Table 4). In addition, the genes that are known to cause causative genes (MAPK1, HNF4A, BMP4, TNF)orwithin syndromic hearing loss in humans are attractive candi- human deafness loci where another gene has been identi- dates for human nonsyndromic hearing loss when they

FIG. 2. The central part of the most significant network generated by IPA network analysis of genes implicated in nonsyndromic or syndromic sensorineural hearing loss (Analysis 2).

Otology & Neurotology, Vol. 34, No. 5, 2013

FIG. 3. An area of the most significant network generated by IPA network analysis of genes implicated in otic capsule development and malformation (Analysis 3). fall within the loci for nonsyndromic deafness (Supple- linked to nonsyndromic sensorineural hearing loss; our mental Tables S1YS6) because several well-studied results strongly support future exploration of this poten- genes, such as MYO7A (15,16) and SLC26A4 (17,18), are tial causal link. TGFB1 is a multifunctional peptide that known to cause syndromic or nonsyndromic hearing loss, controls proliferation, differentiation, extracellular matrix depending on the location and severity of a mutation. production, adhesion, migration, and other functions in many cell types (21). In the inner ear, TGFB1 stimulates DISCUSSION otic capsule chondrogenesis during early development (22) and controls remodeling of the developing otic capsule by A comprehensive network and pathway analysis of inhibiting osteoclast activation and stimulating osteoblast the human deafness genes has revealed that these genes differentiation (23). Our results suggest that TGFB1 may form a small and well-interconnected neighborhood in play additional, as of yet unrecognized, roles in the phy- the cacophony of the human interactome, as reflected in siology of the cochlear sensorineural structures, in addition the very highly significant top networks for each of the 3 to its established role in otic capsule development and re- analyses that we performed. Using musical analogy, a modeling. Interestingly, BMP4, which we have identified node is a note or a chord, a pathway is an instrumental as another candidate gene for human deafness (Table 1), tune linking interconnected notes and chords, and a net- is a member of the TGFB superfamily of signaling work is a symphony involving interconnected tunes from molecules (24). many different instruments. In this harmony of sound, Beta-estradiol is another prominent node in the analy- TGFB1 stands out as the only note of central significance ses of nonsyndromic and syndromic sensorineural hearing in all 3 analyses (Table 1). Although mutations in this loss (Table 1). This molecule is a sex hormone derived gene are known to cause Camurati-Engelmann disease from cholesterol, present not only in women but also in (19) and a coding polymorphism in TGFB1 is associated men because it is derived from testosterone by the action of with otosclerosis (20), TGFB1 has not been causally aromatase. In addition to its critical role in reproduction

Otology & Neurotology, Vol. 34, No. 5, 2013

TABLE 3. Chromosomal location and expression of the nodal molecules summarized in Table 1 IPA Entrez gene Entrez gene ID Location in Deafness locus Entrez gene ID Location in Expression in symbol name (74) for human (74) human (74) in human (6) for mouse (74) mouse (74) mouse (12,13) TGFB1 Transforming growth factor, 7040 19q13.1 DFNA4- 21803 7 6.5 cM beta 1 beta-estradiol^ V Cytochrome P450, family 19, 1588 15q21.1 DFNB16- 13075 9 31.0 cM H subfamily a, polypeptide 1(CYP19A1) F Actin+ V ERK1/2 Mitogen activated protein 5594 22q11.21 DFNB40** 26413 16 9.82 cM H, N kinase 1, (MAPK1) collagen+ V HGF Hepatocyte growth factor 3082 7q21.1 DFNB39 15234 5 4.0 cM (hepapoietin A; scatter factor) Gpcr+ V BMP4 Bone morphogenetic 652 14q22Yq23 DFNA23** 12159 14 15.0 cM H, N protein 4 HNF4A Hepatocyte nuclear factor 3172 20q13.12 DFNB65** 15378 2 94.0 cM N 4, alpha DFNA13- DFNA31** TNF Tumor necrosis factor 7124 6p21.3 DFNB53- DFNB66- 21926 17 19.06 cM H DFNB67- Entries with ‘‘+’’ indicate families of proteins. Genes highlighted in bold represent novel candidates for deafness genes because they are located within deafness loci that await identification of causative genes (**) or within deafness loci where another gene has been identified as causative (-). Beta estradiol is marked with ‘‘^’’ because it is a steroid, not encoded by a gene. However, its conversion from testosterone is mediated by the highlighted gene CYP19A1. ^: Among the listed molecules, only Hgf is known to cause deafness in the mouse (nonsyndromic) (75). Expression refers to mouse cochlear hair cells (H)12 or cochlear neurons (N)13. and sexual function, beta estradiol has an osteoprotective G protein coupled receptors represent the central node (25), neuroprotective (26), and otoprotective role (27). It of the second most significant network in the analysis of is interesting that the gene encoding aromatase, called nonsyndromic and syndromic sensorineural deafness CYP19A1, is located on chromosome 15q21.1, which falls (Table 1). GPCR are a large family of proteins, including within the DFNB16 deafness locus (Table 3). This strongly approximately 800 different members divided into 5 sub- motivates future studies of the role CYP19A1 plays in families, all located within the plasma membrane. They are hearing and deafness. activated by many different ligands, conveying the vast Another key node is ERK1/2 (Table 1)Vachord majority of signal transduction across cell membranes, and with2notes:ERK1andERK2.ERK1/2aresignaling regulating diverse physiologic processes (34,35). GPCR fa- kinases that are members of the mitogen-activated pro- mily has been associated with deafness, as mutations of its tein kinases (MAPKs) with contrasting function, depending largest member, VLGRI, are responsible for Usher syndrome, on stimulus: ERK1/2 activation promotes cell proliferation type 2C (36). Many GPCR genes are novel candidates for and survival (28,29), or cell death and apoptosis (30). deafness genes (Table 4 and Supplemental Table S4). Although MAPKs can be activated by a broad variety of When analyzing genes involved in otic capsule deve- stimuli, it is interesting that ERK 1/2 are frequently activated lopment and malformation, the central node of the second by GPCR (31)Vanother key node in the analysis of syn- most significant network was HNF4A whoseroleinhearing dromic and nonsyndromic sensorineural deafness (Table 1). has not been previously studied. HNF4A is a transcription Although mutations in genes encoding ERK1 and ERK2, factor mainly expressed in the liver but also in the small called MAPK3 and MAPK1, respectively, have not been intestine, colon, kidney, and pancreatic beta cells (37,38). implicated in human deafness, MAPK1 is located within the HNF4A is critically involved in the regulation of the ex- DFNB40 locus (Table 3). Our results motivate future studies pression of many hepatic genes; its absence causes meta- of MAPK1 as a possible causative gene for DFNB40 non- bolic dysfunction and death (39Y41). In addition, HFN4A syndromic, autosomal recessive, profound, prelingual deaf- regulates the expression of genes encoding coagulation ness. Our results also motivate future studies of ERK1/2 factors XII and XIIIb (42). Heterozygous mutations in signaling in otic capsule development and function (Table 1). HNF4A are associated with maturity-onset diabetes of the Animal models point to ERK1/2 being important for young, Type 1 (MODY-1), characterized by abnormal hearing. Inhibition of ERK 1/2 activation causes loss of secretion of insulin by the B cell of pancreas after glucose outer hair cells and enhancement of gentamicin ototoxi- stimulation, with normal liver function and insulin sensi- city in cochlear explants (32). In contrast, ERK 1/2 acti- tivity (43). Hearing has not been studied in people with vation frequently signals damage in the neonatal rat cochlea, MODY-1. Chromosomal location of HNF4A on 20q13.12 and neomycin-induced ERK 1/2 activation in supporting within the DFNB65 locus makes HNF4A an attractive cells promotes hair cell death (33). candidate gene for human deafness (Table 3).

Otology & Neurotology, Vol. 34, No. 5, 2013

TABLE 4. Additional candidate genes for human deafness based on their presence in the networks studied (Table 1, and Supplemental Tables S2Y26), unreported association with human deafness (Supplemental Table S1), and chromosomal location within deafness loci whose causative genes remain unidentified (**) Entrez gene Entrez gene ID for Location in Deafness locus Entrez gene ID Expression in IPA symbol name (74) human (74) human (74) in human (6) for mouse (74) mouse (12,13) ATP6V1B1 ATPase, H+ transporting, 525 2p13.1 DFNA58** 110935 H lysosomal 56/58kDa, V1 subunit B1 BMP6 Bone morphogenetic 654 6p24-p23 DFNA21** 12161 H protein 6 CA14 Carbonic anhydrase XIV 23632 1q21 DFNA7** DFNA49** 23831 H, N CALML4 Calmodulin-like 4 91860 15q23 DFNB48** 75600 H, N CALCA Calcitonin-related 796 11p15.2 DFNA32** H, N polypeptide alpha CLDN10 Claudin 10 9071 13q31-q34 DFNA33** 58187 H, N CNST Consortin, connexin 163882 1q44 DFNB45** 226744 H, N sorting protein ELOVL2 ELOVL fatty 54898 6p24.2 DFNA21** 54326 H, N acid elongase 2 EPS8L2 EPS8-like 2 64787 11p15.5 DFNA32** 98845 H GPR33 G protein-coupled 2856 14q12 DFNA9- DFNA53**, DFNB5** H, N receptor 33 (gene/pseudogene) GPR85 G protein-coupled 54329 7q31 DFNB4- 64450 H, N receptor 85 DFNB14**DFNB17** GPR141 G protein-coupled 353345 7p14.1 DFNB44** 353346 H receptor 141 GPR155 G protein-coupled 151556 2q31.1 DFNB27**DFNB59- 68526 H, N receptor 155 GJA1 Gap junction protein, 2697 6q21-q23.2 DFNB38** 14609 H, N alpha 1, 43kDa ITGA10 Integrin, alpha 10 8515 1q21 DFNA7** 213119 H, N DFNA49** LDHA Lactate dehydrogenase A 3939 11p15.4 DFNA32** 16828 H, N LGR4 Leucine-rich repeat 55366 11p14-p13 DFNB18- 107515 H, N containing G protein-coupled receptor 4 DFNB51** MFHAS1 Malignant fibrous 9258 8p23.1 DFNM2** 52065 H, N histiocytoma amplified sequence 1 MYCN v-myc myelocytomatosis 4613 2p24.3 DFNB47**DFNB83** 18109 H, N viral related oncogene, neuroblastoma derived (avian) PLIN2 Perilipin 2 123 9p22.1 DFNA47** 11520 H, N PTPN11 Protein tyrosine 5781 12q24 DFNA41**DFNA64- 19247 H phosphatase, non-receptor type 11 PRDX2 Peroxiredoxin 2 7001 19p13.2 DFNA57** DFNB15- 21672 H, N DFNB68**DFNB72- DFNB81** SLC39A1 Solute carrier family 39 27173 1q21 DFNA7** 30791 H,N (zinc transporter), member 1 DFNA49** SMARCA4 SWI/SNF-related, 6597 19p13.2 DFNA57**DFNB15- 20586 H,N matrix-associated, actin-dependent regulator of chromatin, subfamily a, member 4 DFNB68**DFNB72- DFNB81** TMEM17 Transmembrane 200728 2p15 DFNA58** 103765 H protein 17 Some of these genes also fall within deafness loci where other causative genes have been identified (-). Mouse orthologs of these genes have been reported in cochlear hair cells (H)12 or cochlear neurons (N)13.

Otology & Neurotology, Vol. 34, No. 5, 2013

Emergence of the proinflammatory cytokine, TNF, as target members of this superfamily (34). Therefore, a candidate gene for deafness is intriguing because di- specific targeting of GPCR most relevant for hearing verse inflammatory conditions, such as otitis media and may be challenging. autoimmune disease, can cause hearing loss (44), whereas Blockage of the ERK1/2 pathway has been of great anti-inflammatory medications, such as steroids (45,46) interest for cancer biology because of the pathway’s and TNF blockers (44), can ameliorate hearing loss. How- known role in cancer aggressiveness. Although power- ever, chronic use of some nonsteroidal anti-inflammatory ful inhibitors of ERK 1/2 pathway have been described medications has been associated with hearing loss, in a sex- with promising results (66,67), the production of specific dependent manner (47,48), highlighting the importance drugs targeting these signaling kinases remains challen- of understanding genetic underpinning of cochlear response ging. Given the reported dual role of ERK1/2 in cochlear to inflammation. physiology and pathophysiology, any therapeutic manip- By exploring topologic characteristics of deafness genes ulation of the ERK1/2 pathway will have to be leveraged in the human interactome, we have also discovered bio- against the risk of worsening or accelerating hearing loss logic processes that provide a global prospective on deaf- with such manipulation. ness (Table 2). Some of these processes provide surprising Interestingly, HNF4A expression is inhibited by cyclos- new insight into mechanisms that may be shared between porine (68), a commonly used drug whose chronic use is the process of hearing and diverse processes such as associated with hearing loss (69). This further highlights melanoma signaling, cardiogenesis, and hepatic fibrosis. a likely important role of HNF4A in hearing. Other sub- Our finding that the canonical pathway for cellular effects stances reported to inhibit HNF4A gene expression and of sildenafil (Viagra) is relevant for sensorineural hearing function include the tumor suppressor P53 (70), cytokines, loss is consistent with the reported sudden hearing loss and the peroxisome-proliferatorYactivated receptor-F coac- from phosphodiesterase-5 inhibitors used to treat erectile tivator 1> (71). Curiously, fasting upregulates and feeding dysfunction (49). downregulates HNF4A expression (72), suggesting future In addition to identifying molecules and pathways of dietary measures to modulate the levels and activity of key relevance for hearing and hearing loss, our analyses HNF4A (73). have predicted drug targets to treat or prevent hearing loss. Although emerging therapies for deafness are likely to benefit from rapid developments of pharmacother- CONCLUSION apies in seemingly unrelated fields, such as cancer biology This study is the first systematic network and pathway or sexual dysfunction, we need to be cognizant of poten- analysis of human deafness genes. By studying genetic tially harmful effects of these therapies on hearing. Many chords, tunes, and symphonies we have identified novel different drugs are known to target the top nodal mole- candidate genes for deafness and highlighted biologic cules of the 5 networks that we analyzed (Table 1). processes that may provide better understanding of hear- Majority of the existing drugs that target TGFB1 focus ing and hearing loss. Our analyses have implications for on enhancing TGFB1 activation to promote chon- design of new drugs and therapies to prevent or treat drogenesis or new bone formation. However, several deafness and for strategies to benefit from parallel devel- studies propose inhibition of gene function using agents opments of pharmacotherapies in other fields of medicine. such as neutralizing antibodies, soluble receptors, recep- A caveat of the study is that it is based on the knowledge tor kinase antagonists, antisense reagents, and specific databases that are always evolving because of new dis- drugs (50). The commonly used drugs that are capable coveries that are integrated into the databases. Therefore, of blocking TGFB1, among their other functions, include our analyses inevitably represent a snapshot in time. antagonists of angiotensin II, Type 1 receptors (e.g., losartan), angiotensin-converting enzyme (ACE) inhibi- Acknowledgments: The authors thank the National Institute tors (51Y56), statins (especially simvastatin) (57Y60), on Deafness and Other Communication Disorders (NIHYNIDCD the oral antimicrobial pirfenidone (61,62) (InterMune K08 DC010419 to K. M. S.), and the Bertarelli Foundation Inc., CA, USA), and the antifibrotic drug tranilast (63) (K. M. S.) for the support. The authors also thank Dr. Joe Adams (Kissei Pharma, Japan). Furthermore, a number of spe- for helpful comments on earlier versions of the article. cific TGFB1 inhibitors are being investigated, including the pan-TGF-b monoclonal antibody (64), the soluble REFERENCES TGF-ARII receptor, antisense oligonucleotides, and inhibitors of ALK5 (50). Interestingly, long-term treatment 1. Traynor B, ‘‘The incidence of hearing loss around the world.’’ with the anti-TGF-b mAb was not found to result in sig- Hearing International. Available at: http://hearinghealthmatters.org/ hearinginternational/2011/incidence-of-hearing-loss-around-the- nificant pathologic changes in major organs and tissues in world/ (2012). Accessed March 1, 2012. mice (64), although other studies have raised concerns 2. Morton NE. Genetic epidemiology of hearing impairment. Ann N regarding the potential acceleration of tumor progression YAcadSci1991;630:16Y31. after TGF-b inhibition (65). 3. Hilgert N, Smith RJ, Van Camp G. Forty-six genes causing nonsyndromic hearing impairment: which ones should be analyzed in Regarding GPCR, which are involved in nearly all DNA diagnostics? Mutat Res 2009;681:189Y96. human physiologic processes, about half of the cur- 4. Bitner-Glindzics M. Hereditary deafness and phenotyping in humans. rently available drugs on the market nonspecifically Br Med Bull 2002;63:73Y94.

Otology & Neurotology, Vol. 34, No. 5, 2013

5. Cryns K, Van Camp G. Deafness genes and their diagnostic appli- 31. Wood K, Sarnecki C, Roberts TM, Blenis J. c-RAS mediates nerve cations. Audiol Neurootol 2004;9:2Y22. growth factor receptor modulation of three signal-transducing pro- 6. Hereditary Hearing Loss Homepage. Available at: http:// tein kinases: MAP kinase, Raf-1 and RSK. Cell 1992;68:1041Y50. hereditaryhearingloss.org/. Accessed March 29, 2012. 32. Battaglia A, Pak K, Brors D, Bodmer D, Frangos JA, Ryan AF. 7. Schrijver I. Hereditary non-syndromic sensorineural hearing loss: Involvement of ras activation in toxic hair cell damage of the transforming silence to sound. J Mol Diagn 2004;6:275Y84. mammalian cochlea. Neuroscience 2003;122:1025Y35. 8. Friedman TB, Griffith AJ. Human nonsyndromic sensorineural deaf- 33. Lahne M, Gale JE. Damaged-induced activation of ERK 1/2 in ness. Annu Rev Genomics Hum Genet 2003;4:341Y402. cochlear supporting cells is a hair cell death-promoting signal that 9. Linghu B, Snitkin ES, Hu Z, Xia Y, DeLisi C. Genome-wide depends on extracellular ATP and calcium. J Neurosci 2008;28: prioritization of disease genes and identification of disease-disease 4918Y28. associations from an integrated human functional linkage network. 34. Rozenfeld R, Devi LA. Exploring a role for heteromerization in Genome Biol 2009;10:R91. GPCR signalling specificity. Biochem J 2010;433:11Y8. 10. Accetturo M, Creanza TM, Santoro C, et al. Finding new genes for 35. Millar RP, Newton CL. The year in G protein-coupled receptor non-syndromic hearing loss through an in silico prioritization study. research. Mol Endocrinol 2010;24:261Y74. PLoS One 2010;5:pii e 12742. 36. McMillan DR, White PC. Studies on the very large G protein- 11. Fisher RA. On the interpretation of x2 from contingency tables, and coupled receptor: from initial discovery to determining its role in the calculation of P. J R Stat Soc 1922;85:81Y94. sensorineural deafness in higher animals. Adv Exp Med Biol 2010; 12. Scheffer D, Shen J, Mingqian Huang, Corey D, Chen ZY. Cell-specific 706:76Y86. gene expression in the inner ear with FACS and RNA-Seq (in pre- 37. Drewes T, Senkel S, Holewa B, Ryffel GU. Human hepatocyte paration). Available at: http://shield.hms.harvard.edu/gene_search.html. nuclear factor 4 isoforms are encoded by distinct and differentially Accessed March 29, 2012. expressed genes. Mol Cell Biol 1996;16:925Y31. 13. Lu CC, Appler JM, Houseman EA, Goodrich LV. Developmental 38. Jiang S, Tanaka T, Iwanari H, et al. Expression and localization of profiling of spiral ganglion neurons reveals insights into auditory P1 promoter-driven hepatocyte nuclear factor-4alpha (HNF4alpha) circuit assembly. J Neurosci 2011;31:10903Y18. isoforms in humans and rats. Nucl Recept 2003;1:5. 14. Kenneson A, Van Naarden Braun K, Boyle C. GJB2 (connexin 26) 39. Hayhurst GP, Lee YH, Lambert G, Ward JM, Gonzalez FJ. Hepa- variants and nonsyndromic sensorineural hearing loss: a HuGE tocyte nuclear factor 4alpha (nuclear receptor 2A1) is essential for review. Genet Med 2002;4:258Y74. maintenance of hepatic gene expression and lipid homeostasis. Mol 15. Weil D, Blanchard S, Kaplan J, et al. Defective myosin VIIA gene Cell Biol 2001;21:1393Y403. responsible for Usher syndrome type 1B. Nature 1995;374:60Y1. 40. Inoue Y, Hayhurst GP, Inoue J, Mori M, Gonzalez FJ. Defective 16. Liu XZ, Walsh J, Mburu P, et al. Mutations in the myosin VIIA ureagenesis in mice carrying a liver-specific disruption of hepatocyte gene cause non-syndromic recessive deafness. Nat Genet 1997; nuclear factor 4alpha (HNF4alpha). HNF4alpha regulates ornithine 16:188Y90. transcarbamylase in vivo. JBiolChem2002;277:25257Y65. 17. Everett LA, Glaser B, Beck JC, et al. Pendred syndrome is caused 41. Inoue Y, Yu AM, Inoue J, Gonzalez FJ. Hepatocyte nuclear factor by mutations in a putative sulphate transporter gene (PDS). Nat 4alpha is a central regulator of bile acid conjugation. J Biol Chem Genet 1997;17:411Y22. 2004;279:2480Y9. 18. Li XC, Everett LA, Lalwani AK, et al. A mutation in PDS causes 42. Inoue Y, Peters LL, Yim SH, Inoue J, Gonzalez FJ. Role of hepa- non-syndromic recessive deafness. Nat Genet 1998;18:215Y7. tocyte nuclear factor 4alpha in control of blood coagulation factor 19. Campos-Xavier AB, Saraiva JM, Savarirayan R, et al. Phenotypic gene expression. J Mol Med 2006;84:334Y44. variability at the TGF-beta-1 locus in Camurati-Engelmann dis- 43. Gupta RK, Vatamaniuk MZ, Lee CS, et al. The MODY1 gene HNF- ease. Hum Genet 2001;109:653Y8. 4alpha regulates selected genes involved in insulin secretion. J Clin 20. Thys M, Schrauwen I, Vanderstraeten K, et al. The coding poly- Invest 2005;115:1006Y15. morphism T263I in TGF-b1 is associated with otosclerosis in two 44. Satoh H, Firestein GS, Billings PB, Harris JP, Keithley EM. Tumor independent populations. Hum Mol Genet 2007;16:2021Y30. necrosis factor-alpha, an initiator, and etanercept, an inhibitor of 21. Janssens K, ten Dijke P, Janssens S, Van Hul W. Transforming cochlear inflammation. Laryngoscope 2002;112:1627Y34. growth factor-beta1 to the bone. Endocr Rev 2005;26:743Y74. 45. McCabe BF. Autoimmune sensorineural hearing loss. Ann Otol 22. Frenz DA, Galinovic-Schwartz V, Liu W, Flanders KC, Van de Rhinol Laryngol 1979;88:585Y9 Water TR. Transforming growth factor beta 1 is an epithelial- 46. Parnes LS, Sun AH, Freeman DJ. Corticosteroid pharmacokinetics derived signal peptide that influences otic capsule formation. Dev in the inner ear fluids: an animal study followed by clinical appli- Biol 1992;153:324Y36. cation. Laryngoscope 1999;109:1Y17. 23. Oreffo ROC, Mundy GR, Seyedin SM, Bonewald LF. Activation of 47. Curhan SG, Eavey R, Shargorodsky J, Curhan GC. Analgesic use the bone-derived latent Tgf beta-complex by isolated osteoclasts. and the risk of hearing loss in men. Am J Med 2010;123:231Y7. Biochem Biophys Res Commun 1989;158:817Y23. 48. Curhan SG, Shargorodsky J, Eavey R, Curhan GC. Analgesic use and 24. Bakrania P, Efthimiou M, Klein JC, et al. Mutations in BMP4 cause the risk of hearing loss in women. Am J Epidemiol 2012;176:544Y54. eye, brain, and digit developmental anomalies: overlap between the 49. Maddox PT, Saunders J, Chandrasekhar SS. Sudden hearing loss BMP4 and hedgehog signaling pathways. Am J Hum Genet 2008; from PDE-5 inhibitors: a possible cellular stress etiology. Lar- 82:304Y19. yngoscope 2009;119:1586Y9. 25. Gennari L, Merlotti D, Nuti R. Aromatase activity and bone loss. 50. Prud’homme GJ. Pathobiology of transforming growth factor beta Adv Clin Chem 2011;54:129Y64. in cancer, fibrosis and immunologic disease, and therapeutic con- 26. Behl C, Widman M, Trapp T, Holsboer F. 17-beta estradiol protects siderations. Lab Invest 2007;87:1077Y91. neurons from oxidative stress-induced cell death in vitro. Biochem 51. Wolf G. Renal injury due to rennin-angiotensin-aldosterone sysyem Biophys Res Commun 1995;216:473Y82. activation of the transforming growth factor-beta pathway. Kidney 27. Hultcrantz M, Simonoska R, Stenberg AE. Estrogen and hearing: a Int 2006;70:1914Y9. summary of recent investigations. Acta Otolaryngol 2006;126:10Y4. 52. Tox U, Steffen HM. Impact of inhibitors of the renin-angiotensin- 28. Xia Z, Dickens M, Raingeaud J, Davis RJ, Greenberg ME. Opposing aldosterone system on liver fibrosis and portal hypertension. Curr effects of ERK and JNK-p38MAPkinases on apoptosis. Science Med Chem 2006;13:3649Y61. 1995;270:1326Y31. 53. Yao HW, Zhu JP, Zhao MH, Lu Y. Losartan attenuates bleomycin 29. Arthur DB, Georgi S, Akassoglou K, Insel PA. Inhibition of apoptosis induced pulmonary fibrosis in rats. Respiration 2006;73:236Y42. by P2Y2 receptor activation: novel pathways for neuronal survival. J 54. Bujak M, Frangogiannis NG. The role of TGF-beta signaling in Neurosci 2006;26:3798Y804. myocardial infarction and cardiac remodeling. Cardiovasc Res 30. Pearson G, Robinson F, Beers Gibson T, et al. Mitogen activated 2007;74:184Y95. protein (MAP) kinase pathways: regulation and physiological 55. Cohn RD, Van Erp C, Habashi JP, et al. Angiotensin II type 1 functions. Endocr Rev 2001;22:153Y83. receptor blockade attenuates TGF-beta-induced failure of muscle

Otology & Neurotology, Vol. 34, No. 5, 2013

regeneration in multiple myopathic states. Nat Med 2007;13: induced tumor formation and metastasis. Cancer Res 2005;65: 204Y10. 2296Y302. 56. Dussaule JC, Chatziantoniou C. Reversal of renal disease: is it 66. Tanimura S, Asato K, Fujishiro SH, Kohno M. Specific blockade of enough to inhibit the action of angiotensin II? Cell Death Differ the ERK pathway inhibits the invasiveness of tumor cells: down- 2007;14:1343Y9. regulation of matrix metalloproteinase -3/-9/-14 and CD44. Bio- 57. Kim SI, Kim HJ, Han DC, Lee HB. Effect of lovastatin on small chem Biophys Res Commun 2003;304:801Y6. GTP binding proteins an on TGF-beta1 and fibronectin expression. 67. Kohno M, Poyssegur J. Pharmacological inhibitors of the ERK Kidney Int Suppl 2000;77:S88YS92. signaling pathway: applications as anticancer drugs. Prog Cell Y 58. Watts KL, Spiteri MA. Connective tissue growth factor expression Cycle Res 2003;5:219 24. and induction by transforming growth factor-beta is abrogated by 68. Borlak J, Niehof M. HNF4alpha and HNF1alpha dysfunction as a simvastatin via a Rho signaling mechanism. Am J Physiol Lung Cell molecular rational for cyclosporine induced posttransplantation Mol Physiol 2004;287:L1323Y32. diabetes mellitus. PLoS ONE 2009;4:e4662. 59. Vieira JM Jr, Mantovani E, Rodrigues LT, et al. Simvastatin 69. Marioni G, Perin N, Tregnaghi A, Bellemo B, Staffieri D, de attenuates renal inflammation, tubular transdifferentiation and Filippis C. Progressive sensorineural hearing loss probably induced by chronic cyclosporin A treatment after renal transplantation for interstitial fibrosis in rats with unilateral uteral obstruction. Nephrol Y Dial Transplant 2005;20:1582Y91. focal glomerulosclerosis. Acta Otolaryngol 2004;124:603 7. 70. Maeda Y, Hwang-Verslues WW, Wei G, et al. Tumor suppressor 60. Redondo S, Santos-Gallego CG, Tejerina T. TGF-beta1: a novel p53 down-regulates the expression of the human hepatocyte nuclear target for cardiovascular pharmacology. Cytokine Growth Factor factor 4> (HNF4>) gene. Biochem J 2008;415:289Y96. Rev 2007;18:279Y86. Y 71. Wang Z, Burke PA. Modulation of hepatocyte nuclear factor-4a 61. Lasky J. Pirfenidone. IDrugs 2004;7:166 72. function by the peroxisome-proliferator-activated receptor-F co- 62. Antoniu SA. Pirfenidone for the treatment of idiopathic pulmonary activator-1> in the acute phase response. Biochem J 2008;415: Y fibrosis. Expert Opin Investig Drugs 2006;15:823 8. 289Y96. 63. Wojtowicz-Praga S. Reversal of tumor-induced immunosuppres- 72. Xie X, Liao H, Dang H, et al. Down-regulation of hepatic sion by TGF-beta inhibitors. Invest New Drugs 2003;2:21Y32. HNF4alpha gene expression during hyperinsulinemia via SREBPs. 64. Ruzek MC, Hawes M, Pratt B, et al. Minimal effects on immune Mol Endocrinol 2009;23:434Y43. parameters following chronic anti-TGF-beta monoclonal antibody 73. Hwang-Veslues W, Sladek FM. HFN4aVrole in drug metabolism administration to normal mice. Immunopharmacol Immunotoxicol and potential drug target? Curr Opin Pharmacol 2010;10:698Y705. 2003;25:235Y57. 74. Online Mendelian Inheritance in Man Database Homepage (OMIM). 65. Forrester E, Chytil A, Bierie B, et al. Effect of conditional knockout Available at: http://omim.org/. Accessed September 19, 2012. of the type II TGF-beta receptor gene in mammary epithelia on 75. The Jackson Laboratory Homepage. Available at: http:// mammary gland development and polyomavirus middle T antigen hearingimpairment.jax.org/. Accessed September 19, 2012.

Erratum

Long-Term Benefit and Sound Localization in Patients With Single-Sided Deafness Rehabilitated With an Osseointegrated Bone-Conduction Device: Erratum

In the article that appeared on page 111 of the January 2013 (Volume 34, Issue 1) issue of Otology & Neurotology, errors have been noted in the author names. They should have appeared as: Nicolas Saroul, Mohamed Akkari, Yoann Pavier, Laurent Gilain, and Thierry Mom.

REFERENCE

Nicolas S, Mohamed A, Yoann P, Laurent G, Thierry M. Long-term benefit and sound localization in patients with single-sided deafness rehabilitated with an osseointegrated bone-conduction device. Otol Neurotol 2013;34:111Y4.

Otology & Neurotology, Vol. 34, No. 5, 2013