A mutation in Nischarin causes otitis media via LIMK1 and NF-κB pathways

Otitis media (OM), inflammation of the middle ear (ME), is a common cause of conductive hearing impairment. Despite the importance of the disease, the aetiology of chronic and recurrent forms of middle ear inflammatory disease remains poorly understood. Studies of the human population suggest that there is a significant genetic component predisposing to the development of chronic OM, although the underlying genes are largely unknown. Using N-ethyl-N-nitrosourea mutagenesis we identified a recessive mouse mutant, edison, that spontaneously develops a conductive hearing loss due to chronic OM. The causal mutation was identified as a missense change, L972P, in the Nischarin (NISCH) gene. edison mice develop a serous or granulocytic effusion, increasingly macrophage and neutrophil rich with age, along with a thickened, inflamed mucoperiosteum. We also identified a second hypomorphic allele, V33A, with only modest increases in auditory thresholds and reduced incidence of OM. NISCH interacts with several proteins, including ITGA5 that is thought to have a role in modulating VEGF-induced angiogenesis and vascularization. We identified a significant genetic interaction between Nisch and Itga5; mice heterozygous for Itga5-null and homozygous for edison mutations display a significantly increased penetrance and severity of chronic OM. In order to understand the pathological mechanisms underlying the OM phenotype, we studied interacting partners to NISCH along with downstream signalling molecules in the middle ear epithelia of edison mouse. Our analysis implicates PAK1 and RAC1, and downstream signalling in LIMK1 and NF-κB pathways in the development of chronic OM.


Introduction
Otitis media (OM) is characterised by inflammation of the middle ear (ME), often associated with a conductive hearing impairment, and is the commonest cause of hearing loss in children. It is perceived by many to be a transient affliction that in reality places a substantial social, medical and economic burden on healthcare systems globally [1]. Evidence from studies of the human population suggests that there is a significant genetic component predisposing to the development of recurrent and chronic forms of OM [2,3]. Despite the importance of the disease, many of the genes involved in OM susceptibility have still yet to be identified. At present, the use of mouse models is the most promising method to identify candidate loci underlying susceptibility to OM. Mouse models have highlighted the role of Toll-like receptors (TLRs) in acute OM, in particular the protection against commensal and pathogenic bacteria, and that persistent NF-κB or TGF-β signalling could be two mechanisms leading to the overactive proinflammatory response seen in chronic OM [4,5].
The large-scale phenotype-driven mouse ENU (N-ethyl-N-nitrosourea) mutagenesis program at MRC Harwell [6,7] has previously identified two novel mouse mutants, Jeff and Junbo, that develop a conductive hearing loss characterised by ME fluid and mucosal inflammation [8,9]. The Jeff mouse has a mutation in the Fbxo11 gene [10] and the Junbo mouse has a mutation in Evi1 [9]. Studies have revealed that these genes, are involved in signalling of the TGF-β superfamily, via SMAD proteins [11,12]; negatively regulate NF-κB-dependent inflammation [13]; and highlight the role of HIF-VEGF pathways in the underlying genetic and pathophysiological mechanisms that predispose to chronic OM [14].
OM mouse models with single gene mutations have identified a number of genes as candidate susceptibility genes for human OM, including Tlr2, Tlr4, p73, E2f4, Plg, Tgif1, Evi1 and Fbxo11 [4]. These genes identified from mouse models of OM are beginning to be studied in the human population; with significant associations between OM and polymorphisms in FBXO11 [15,16], TLR2 [17] and TLR4 [17][18][19].
We have identified and characterised a novel OM mouse mutant, edison, from the ENU mutagenesis program at MRC Harwell. Homozygous edison mice spontaneously develop a conductive hearing loss associated with chronic inflammation of the ME, sharing many features with chronic OM in humans. The underlying mutation of this phenotype has been identified as a mutation in the Nisch gene. We have explored the role of Nisch in chronic OM, relating the edison phenotype to the underlying mechanisms of Nisch function. We have utilised double mutants to assess genetic interactions and pathways involved, implicating PAK1 and RAC1, and downstream signalling events in LIMK1 and NF-κB signalling pathways in the development of chronic OM. Overall, the edison mouse highlights a new candidate gene for susceptibility to chronic OM and has provided further insight into the genetic pathways and pathogenic processes involved.

Identification of the edison mutation
A phenotype-driven ENU mutagenesis screen [20] identified a new recessive mutant, edison (edsn), with hearing loss. Preliminary phenotyping using a click-box test (20 kHz, 90 dB sound pressure level (SPL) tone burst) of an age-matched cohort derived from the founder mouse indicated that 6-week-old (wk) mice demonstrated a reduced startle response. SNP-based linkage analysis and mapping identified an approximately 9 Mb interval on chromosome 14 delineated by marker rs30778552 and rs46823676 containing 119 genes ( Fig 1A). Whole-genome sequencing identified 60 ENU-induced, de novo variants within the critical 9Mb interval and importantly only one missense variant was identified. This missense variant was a c.3079T>C substitution in Nischarin (Nisch) (open reading frame of NCBI RefSeq transcript NM_022656.2) that results in a Leu972Pro substitution ( Fig 1B). The change occurs in a highly conserved region that has been maintained through evolution ( Fig 1C). PROVEAN analysis predicts that this change is ''deleterious" and SIFT predicts that it is ''not tolerated". No other non-synonymous sequence changes were identified within the minimal interval. The Nisch locus encodes a protein of 1,593 amino acids, coded for by 22 exons. The protein consists of an N-terminal phox homology (PX) domain, six putative leucine-rich repeats (LRRs), a predicted coiled-coil (CC) domain, an alanine/ proline-rich region and a long C-terminal region (Fig 1D).
DNA and sperm archives derived from ENU mutagenesis programmes [21] were utilised to identify an additional allele at the Nisch locus. We screened ten exons of Nisch employing high resolution melting analysis of~10,000 mutant mice and identified a c.98T>C substitution resulting in a Val33Ala substitution within a conserved region of the NISCH PX domain. We rederived this second allele, Nisch V33A , and examined the phenotypes.
Mild craniofacial defects in edison mice. We found that all Nisch edsn/edsn mice demonstrated a shortened snout compared with wild-type mice, indicating a mild craniofacial abnormality (S1 Fig). To investigate the craniofacial phenotype of mice, we took dorsoventral X-ray images of the skulls and computed cranial measurements (S1C Fig). To determine whether any of the cranial bones showed disproportionate growth, allometric comparisons against skull length were used to normalise the data for differences in body size (S1C and S1D Fig). Evaluation of Nisch edsn/edsn and wild-type skulls revealed a significant difference in allometric comparisons between nasal bone/skull length (Kruskal-Wallis, p < 0.001) and frontal bone/ skull length (Kruskall-Wallis: p = 0.036). The magnitude of the difference between Nisch edsn/edsn and wild-type mice was small (~3%). No other skeletal defects were evident.
To investigate the orientation and morphology of the Eustachian tube (ET) in edison mice, the skulls were stained and the angles between the midline of the skull and the bony part of the left and right ET were measured (S1E and S1F Fig). No statistically significant differences in ET measurements were detected between Nisch edsn/edsn and wild-type mice (Kruskall-Wallis: Left, p = 0.848; Right, p = 0.324).
Hearing deficits in Nisch edsn/edsn animals due to spontaneous otitis media. In order to assess the onset and progression of the hearing impairment in edison mice, click-evoked ABR tests to measure auditory function were performed in cohorts of age matched littermates from 3 wk to 20 wk (Fig 2A). Nisch +/+ and Nisch edsn/+ mice showed normal ABR thresholds throughout the time course, whereas significantly elevated thresholds were detected in Nisch edsn/edsn mice as early as 3 wk and progressively increased with age (mean ABR threshold: Nisch edsn/edsn , 3 wk = 30 ± 1 dB SPL, n = 36; 20 wk = 46 ± 2 dB SPL, n = 61; Kruskall-Wallis: p < 0.001). We showed that the mean ABR thresholds were elevated by about 20-30 dB SPL in Nisch edsn/edsn animals when compared with wild-type mice, suggesting a conductive hearing loss.
Visual inspection of the tympanic membrane is a simple method for diagnosing OM. The cloudy appearance of an eardrum is a semi-quantitative measure of bulla fluid accumulation and the incidence of OM was assessed in edison mice ( Fig 2D). Nisch +/+ and Nisch edsn/+ animals showed a small incidence of unilateral OM (Nisch +/+ , 1% unilateral OM, n = 148; Nisch edsn/+ , 2% unilateral OM, n = 194). In Nisch edsn/edsn animals there was a clear trend showing increased prevalence of OM from 3 wk onwards (Fig 2E). At 3 wk, no Nisch edsn/edsn mice had bilateral OM, 40% had unilateral OM and 60% had no OM phenotype (Fisher Exact: p = 0.028). The prevalence of OM progressively increased throughout the time course, where at 20 wk 56% of Nisch edsn/edsn mice had bilateral OM, 40% had unilateral OM and only 4% displayed no OM phenotype (Fisher Exact: p < 0.001).
In Nisch V33A/V33A animals, examination of auditory function revealed that mice develop a milder hearing deficit compared to Nisch edsn/edsn mice ( Fig 2B). Although progressive, the onset of the hearing deficit was observed at 8 wk in Nisch V33A/V33A animals. Visual inspection of the tympanic membrane was used to assess the incidence of OM in Nisch V33A/V33A mice ( Fig  2F). At 12 wk, Nisch V33A/V33A mice exhibited significantly increased prevalence of OM compared to wild-type littermates (Fisher Exact: p = 0.011). In Nisch V33A/V33A mutants, 10% had bilateral OM, 50% had unilateral OM and only 40% displayed no OM phenotype (n = 10). The less significant hearing deficit and limited disease progression suggests that the Nisch V33A allele is hypomorphic in nature. Compound heterozygotes carrying both Nisch edsn and Nisch V33A alleles showed a similar phenotype to Nisch edsn/edsn mice (Fig 2C and 2G). Nisch V33A/edsn mice develop a spontaneous chronic OM, with an associated elevation in ABR hearing thresholds. The hearing deficit was progressive, with onset recorded at 4 wk in Nisch V33A/edsn mice. At 12 wk, the incidence of OM in Nisch V33A/edsn mice was assessed and 50% had bilateral OM, 33% had unilateral OM and only 17% displayed no OM phenotype (n = 6) ( Fig 2G). This increased prevalence of OM in Nisch V33A/edsn mice at 12 wk was significantly different compared to wildtype littermates (Fisher Exact: p = 0.015).
We assessed whether there was a sensorineural element to the hearing loss in Nisch edsn/edsn mice by evaluating frequency-specific auditory function and the inner ear morphology (S2 Fig). We tested 5 mice for each genotype from 4 wk to 20 wk for ABR response at 3 frequencies 8, 16, and 32 kHz. Parallel shifts in audiometric profiles across frequencies at the different ages were recorded, consistent with an underlying conductive hearing loss (S2A Fig). We analysed the inner ear morphology from 5 mice for each genotype at 20 wk to examine the structure of the organ of Corti and the sensory cells (S2B- S2D Fig). Histological examination of mid-modiolar sections of the cochlea revealed no obvious abnormalities in Nisch edsn/edsn mice. Similarly, ultra-structural analysis of the inner ear with scanning electron microscopy showed no evidence of hair cell damage or hair cell loss, with Nisch edsn/edsn mice displaying normal cell and bundle morphology throughout the cochlear turn (basal, mid and apical).
Histological analysis of middle ear demonstrates chronic otitis media. Histological examination was performed to investigate the pathological changes in the ME of edison mice (Fig 3). Wild-type and Nisch edsn/+ mice at 20 weeks displayed clear ME cavities, lined with a thin mucoperiosteum (Fig 3A and 3B). Nisch edsn/edsn mice develop chronic otitis media with ME cavities filled with a cellular exudate and lined with a thickened mucoperiosteum (Fig 3C  and 3D). The mucosal inflammation was diffuse and of moderate severity. Papillary to polypoid exophytic growths were observed projecting into the ME cavity ( Fig 3E). The polypoid projections were most likely due to fibroblast stimulation as a consequence of chronic inflammation. There were small accumulations of inflammatory cells and dilated lymphatic and blood vessels in the mucoperiosteum (Fig 3E). In middle ears with the most severe phenotype, there were instances showing inflammation of the tympanic membrane ( Fig 3F). There were differing types of ME effusion: a thin watery effusion, a dense granulocyte-rich effusion, or a combination of both (Fig 3G-3I). F4/80 stained sections identified increased macrophage infiltration in the exudates of some Nisch edsn/edsn mice, with the presence of enlarged foamy macrophages containing coarsely granular material (Fig 3J).
To further investigate the ME phenotype in edison, heads were sectioned at 3, 4, 6, 8 and 20 wk. At 3 wk, most Nisch edsn/edsn animals displayed a normal wild-type phenotype with clear ears (71%, n = 14 ears), and only three mice had a thin watery effusion (21%, n = 14 ears). At 4 wk, there were five Nisch edsn/edsn ears that contained fluid in the ME cavity with a thickened epithelial lining (50%, n = 10 ears). Two ears had a thin watery effusion, and three ears displayed a cellular effusion with the presence of polymorphonuclear cells (PMNs) that are the first responding inflammatory cells that migrate towards the site of inflammation. In 6 wk Nisch edsn/edsn animals, four ears showed inflammation in the ME cavity with a thickened epithelial lining (33%, n = 12 ears), with three of these ears also containing a cellular effusion. By 8 wk, the inflammation in Nisch edsn/edsn animals had progressed to a chronic inflammation with effusion, as evidenced by a thick cellular fluid in the bulla cavity along with a greatly thickened ME mucosa (67%, n = 12 ears). There was variability in the type of effusion found, where two ears exhibited a serous effusion and in six ears the ME fluid was thick with the presence of macrophages and PMNs. Nisch edsn/edsn mice studied at 20 wk displayed a chronic inflammation with effusion, similar to what was observed at 8 wk. Seventeen ears presented a chronic inflammation (77%, n = 22 ears), eleven of which contained fluid. Ten ears exhibited a thick effusion with the presence of macrophages and PMNs and in only one ear a serous exudate was observed. Cellular debris-like aggregates were present in the thick effusions and developing polyps were observed in the ME mucosa. None of the wild-type or Nisch edsn/+ mice evaluated displayed signs of OM pathology at any time point.
To quantify the chronic inflammation observed in Nisch edsn/edsn mice and assess any subclinical changes in pathology, blinded assessment of the mucoperiosteum thickness was carried out in all edison genotypes from 3 wk to 20 wk ( Fig 3K). As there was variation in the penetrance of OM in Nisch edsn/edsn mice, the ears were categorised into those that had an OM phenotype and those that did not, Nisch edsn/edsn OM and Nisch edsn/edsn Normal respectively. The criteria used to impartially categorise the ears into Nisch edsn/edsn OM and Nisch edsn/edsn Normal was visualisation of tympanic membrane. Nisch edsn/edsn Normal ears displayed normal, wild-type ME mucosal morphology of between 9.0 μm to 17.0 μm mucosal thickness, with no significant differences compared to wild-type littermates. In contrast, mucosal thickness observed in Nisch edsn/edsn OM ears showed a mucosal morphology that progressively thickened with age. At 3 wk, Nisch edsn/edsn OM ears had a mean mucosal thickness of 28.4 ± 8.49 μm and by 20 wk, the mean mucosal thickness had increased to 80.0 ± 7.73 μm. Nisch edsn/edsn OM ears showed significantly increased thickening of the ME mucosa at every time point, when compared to Nisch +/+ ears (Kruskall-Wallis: 3 wk, p = 0.013; 4 wk, p = 0.001; 6 wk, p = 0.06; 8 wk, p < 0.001; 20 wk, p < 0.001).
Upregulation of hypoxia and inflammatory genes in middle ear fluid. Venous blood and ear fluid was used to perform RT-qPCR to look for expression changes in inflammatory and hypoxia responsive genes (S3 Fig). Relative to a normoxic baseline control of Nisch edsn/edsn white blood cells (WBC), the inflammatory cells that accumulate within the bulla fluids of Nisch edsn/edsn mice showed elevated expression of Hif1a (30-fold; t-test: p < 0.001) and the HIF responsive gene Vegfa (87-fold; t-test: p < 0.001). Il1b (15-fold; t-test: p = 0.003) and Tnfa (66-fold; t-test: p < 0.001) are known modulators of Hif1a translation and expression data indicated that they were also elevated in Nisch edsn/edsn bulla fluid inflammatory cells relative to WBC (S3 Fig). Additionally expression of Src, which is a known mediator of VEGF signalling, was also significantly elevated in Nisch edsn/edsn bulla compared to WBC (24-fold; t-test: p = 0.005). Finally, Evi1 and Fbxo11 were expressed in the bulla fluid of Nisch edsn/edsn mice, but only Evi1 (256-fold; t-test: p = 0.006) was expressed at higher levels relative to WBC.
Lung abnormalities. Due to the reduced numbers of Nisch edsn/edsn progeny at weaning age, we studied the mice at E16.5, E18.5 and P0 to search for any physiological defect that may contribute to lethality (Fig 4A-4D). In some Nisch edsn/edsn mice we observed morphological lung defects including thickened interstitial mesenchyme and smaller airspaces than in wildtype littermates. At E16.5, 33% of Nisch edsn/edsn lungs appeared severely affected (n = 6). This was replicated at E18.5 and P0, whereby 30% (n = 10) and 29% (n = 7) of Nisch edsn/edsn lungs appeared severely affected respectively. At each time point, the width of airspaces in severely affected Nisch edsn/edsn mutants was significantly smaller than wild-type tissue (one-way ANOVA: p < 0.001), whereas no difference in the number of airspaces was observed. The reduced numbers of Nisch edsn/edsn progeny are likely a consequence of their lung defect.
Macrophages are the primary phagocytic cells in the lungs. In wild-type lungs, F4/80-positive cells are generally relatively small and rounded ( Fig 4N). In contrast, abnormal accumulations of enlarged alveolar macrophages containing coarsely granular material were observed within the lungs of Nisch edsn/edsn mice ( Fig 4O).
IHC analysis of NF-κB p65 expression showed nuclear localisation and raised levels of protein in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice compared to wild-type (oneway ANOVA: p = 0.017 and p <0.001 respectively). Moreover, there was a significant enhancement of NF-κB labelling in the double mutant compared to Nisch edsn/edsn (one-way ANOVA: p = 0.026) (Fig 6F). In agreement with IHC, we found significantly higher levels of NF-κB p65 by western blot in both mutants compared to wild-type (t-test: p = 0.039 and p = 0.041) ( Fig  7F). We also examined levels of activated NF-κB p65 by western blot employing an antibody that recognises phosphorylated-NF-κB [Ser 276] p65 (p-NF-κB p65) and detected significantly higher levels of the activated protein in both mutants compared to wild-type (t-test: p = 0.030 and p = 0.035) (Fig 7G).
In addition, we investigated the expression of two pathways that are implicated in the development of chronic OM (see Introduction), or are regulated by NISCH interacting partners. First, we evaluated levels of focal adhesion kinase (FAK) and SRC. Cross-talk between integrins and VEGF is a critical factor in the regulation of angiogenesis, vascularisation and vascular permeability [25]. Integrins regulate VE-cadherin via the activation of SRC, and FAK catalytic activity is required when α5β1 integrin stimulates SRC activation through FAK phosphorylation. FAK inhibition prevents VEGF-stimulated vascular permeability underlining the importance of FAK activity in the regulation of adherens junctions [32]. Tyr 397 in human FAK becomes phosphorylated upon integrin engagement and creates a binding site for SRC. This results in release of the inactive conformation of SRC (Tyr 527) and leads to autophosphorylation of SRC on Tyr 416. Activated SRC further phosphorylates FAK on additional residues, one of which is Tyr 576. The activated FAK-SRC complex then initiates multiple downstream signalling pathways [33,34]. IHC of ME epithelial cells showed there was increased expression of total FAK in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn compared to wild-type mice (one-way ANOVA: p = 0.015 and p < 0.001 respectively). A significant difference was also observed between Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn ears (one-way ANOVA: p = 0.005) (Fig 6G).
To study protein levels of FAK we used an antibody raised against the last 50 amino acids at the C-terminal of the human protein. There are nine known mouse isoforms of FAK (http:// www.uniprot.org/uniprot/P34152) produced by alternative promoter usage and alternative splicing and the antibody is potentially able to detect six of them. We detected three main bands in ME epithelial cell lysates at 124, 115 and 100kDa. The full-length canonical isoform is 124kDa. We detected significantly increased levels of the full length 124kDa FAK1 protein in Nisch edsn/edsn mutants compared to wild-type (t-test: p = 0.014). The level of the 115kDa form was significantly raised in Nisch edsn/edsn (t-test: p = 0.014) and Itga5 tm1Hyn/+ ; Nisch edsn/edsn (ttest: p = 0.001) mice compared to wild-type, while the 100kDa form was the main isoform detected in wild-type samples. A significant difference was detected in the levels of the 100kDa between wild-type and Itga5 tm1Hyn/+ ; Nisch edsn/edsn (t-test: p = 0.0005) (Fig 7H). Furthermore we used phosphorylated-FAK (p-FAK) [Y576] and phosphorylated-SRC (p-SRC) [Y527] antibodies to test the activity of the FAK-SRC complex in middle ear epithelia. Using a p-FAK [Y576] antibody, which recognises activated FAK, we detected an increase in the activated FAK in the Nisch edsn/edsn tissue compared to the wild-type ME epithelial cell lysate (t-test: p = 0.007) (Fig 7I). Not surprisingly, in the double mutant Itga5 tm1Hyn/+ ; Nisch edsn/edsn , levels of activated FAK were not significantly different to wild-type. Using a p-SRC [Y527] antibody, which recognises inactive SRC, we detected complementary results to that seen with activated FAK. We found reduced levels of protein in Nisch edsn/edsn ME lysates compared to the wild types (t-test: p = 0.002) but no differences between wild-type and Itga5 tm1Hyn/+ ; Nisch edsn/edsn (Fig 7J).
IHC analysis of this suite of proteins in airway epithelia revealed many similarities to that seen in ME epithelia (S5 Fig). Levels of NISCH appeared to be raised in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice compared to wild-type. Furthermore, increased epithelia expression of p-LIMK1/2 in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice mirrored the findings in ME epithelia. Similarly, RAC1 epithelial levels were significantly raised in Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice, while NF-κB and FAK expression was raised in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice. However, we observed no significant changes in p-SMAD2 levels in mutant airways. In contrast, in lung tissue we observed very few significant changes in protein levels by western blot analysis (S5 Fig). These results may reflect the complexity of tissues and cell types isolated from dissected material and the visible mesenchymal expression for many of these proteins in airway tissue.

Discussion
In a large-scale ENU mutagenesis screen we recovered a new recessive mouse model of chronic OM, edison. The edison mutant carries a Leu972Pro change in the Nischarin gene. Nisch edsn/edsn homozygotes display a progressive middle ear disease with 56% of mice displaying bilateral OM by 20 weeks and elevated ABR thresholds of 20-30 dB SPL indicative of a conductive hearing loss. We derived an additional ENU allele (Nisch V33A ) in the Nischarin gene which also presents with progressive chronic OM, but where ABR thresholds were only very moderately increased and at 12 weeks only 10% of the mice had bilateral OM. It appears that the Nisch V33A allele is severely hypomorphic. Compound heterozygotes of the edison and Nisch V33A alleles showed an intermediate non-complementing OM phenotype. We did not identify any sensorineural element to the hearing loss in the Nisch edsn/edsn mutant.
The chronic OM in Nisch edsn/edsn mice is exemplified by exudate within the ME cavity and a thickened mucoperiosteum and polypoid exophytic growths, sometimes associated with an inflamed tympanic membrane. Serous or granulocyte-rich effusions were observed, but as the mice aged a thick effusion was predominately observed rich in macrophages and PMNs. Examination of ME exudates for upregulation of both inflammatory genes and hypoxia genes found that both Il-1b and Tnfa were both upregulated, as were Hif1a and the HIF responsive gene Vegfa. This is similar to the findings reported for other mouse models of chronic OM, such as the Jeff [8], Junbo [9] and Tgif1 [35] mutants. Both the middle ear and the lungs have substantial similarities in structure and function [36] and, intriguingly, we identified a lung defect in Nisch edsn/edsn mice. During embryonic development, we found a significant reduction in airspace width, while in the adults, we observed an emphysema-like phenotype with enlarged airspace width and a reduced number of airspaces. In utero, the network of airways is generated first followed by formation of the gas-exchanging units (alveoli) that develop from the distal ends of the small airways. As a consequence disruption to lung development often results in narrower airspaces in embryonic lungs but enlarged airspaces post-natally because of insufficient generation of alveoli. These lung abnormalities likely account for the deficit of Nisch edsn/edsn mice recovered from the various crosses that we report.
The discovery of the involvement of Nischarin in the development of inflammatory middle ear disease identifies a novel gene and associated pathways that are involved in OM. This led us to explore the intersection with known pathways of OM [11,13,14] and the downstream signalling mechanisms that lead to the OM phenotype. Nischarin is a highly conserved protein across mammalian species, consisting of an N-terminal phox homology (PX) domain, 6 putative leucine-rich repeats, a coiled-coil domain, an alanine/proline-rich region and a long C-terminal region. NISCH has a multitude of interacting partners, including ITGA5 [37], PAK1 [26], Rac1 [27,38], LIMK1 [28], Rab14, PI3P [38], and LKB1 [39]. Association of NISCH with these interacting partners underlines its broad impact on the regulation of cell motility, cell invasion, vesicle maturation, as well as its role as a tumour suppressor [28,37,38,40]. Most notably, the binding of NISCH to ITGA5 [37] is thought to mediate the translocation of ITGA5 from the cell membrane to endosomes [24] thus regulating ITGA5 levels. As we discuss above, integrins have been shown to play a critical role in modulating VEGF-induced angiogenesis and vascularization [25]. These pathways thus intersect with the hypoxia-response pathways mediated by HIF-1a which have been demonstrated to be involved with the development of chronic OM in the Junbo and Jeff models [14]. Hypoxia leads to upregulation of VEGFA and downstream pathway genes resulting in VEGF-induced angiogenesis and vascular leak. VEGFR inhibitors moderated angiogenesis and lymphangiogenesis in the Junbo mouse. For these reasons we sought to explore the role of ITGA5 in the edison mutant, and also to characterise the responses of downstream pathways which may illuminate the mechanisms of OM development in edison.
We found a strong genetic interaction between edison and Itga5 mutants. Itga5 tm1Hyn/+ ; Nisch edsn/edsn double mutants compared to Nisch edsn/edsn mice showed significantly elevated ABR thresholds as well as a very significant raised frequency of bilateral OM in mice from 4 weeks onwards. In summary, the OM was more highly penetrant from an earlier age, and commensurately less progressive. Given the interactions of ITGA5 and NISCH, along with the interactions of NISCH with diverse molecules involved in LIMK1 and NF-κB signalling, we sought to interpret this genetic interaction in the context of known signalling pathways and interactions downstream of NISCH. Moreover, in developing a mechanistic model, we took into account the reports that interaction of ITGA5 with NISCH affects some of the downstream interactions of the NISCH molecule itself.
RAC1 signalling regulates disparate cellular functions mediated through a variety of effector proteins [30,31]. PAK1 is a key downstream effector of RAC1 [41,42], with binding of RAC1 leading to activation of PAK1 [43]. NISCH represses this pathway, and NISCH has been shown to block RAC1 induced cell migration through binding to PAK1; interaction with NISCH strongly inhibits the kinase activity of PAK1 [23]. RAC1 activation of PAK1 enhances the interaction between NISCH and PAK1, while notably expression of ITGA5 also increases the association between NISCH and PAK1 [23]. LIMK1 is a downstream effector of PAK1 [44] and has been shown to be involved in vascular permeability [45]. NISCH is also an interacting partner with LIMK1, regulating cell invasion through repression of the LIMK1-cofilin pathway [28]. NISCH also regulates PAK1-independent RAC1 signalling through direct interaction with RAC1 [27]. RAC1 stimulates the phosphorylation and degradation of IkB and up-regulates NF-κB [46,47]. Overexpression of NISCH has been shown to suppress the ability of RAC1 to stimulate NF-κB activation [27]. The Junbo OM mutant carries a mutation in the Evi1 gene, and it is noteworthy that EVI1 is a negative-feedback regulator of NF-κB [13]. The mutation in Junbo leads to activation of NF-κB and inappropriate regulation of the inflammatory response [13].
We have assessed the expression of the critical genes within these pathways in the ME epithelia of wild-type, Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice. First, we observed that levels of ITGA5 were raised in Nisch edsn/edsn mice though not significantly, reflecting the role of NISCH in regulating ITGA5 levels. Not surprisingly, ITGA5 was significantly reduced compared to edison mice in the double mutant, Itga5 tm1Hyn/+ ; Nisch edsn/edsn . We found significantly raised levels of activated p-PAK1 in both mutant lines compared to wild-type, which was mirrored by downstream raised levels of p-LIMK1/2 in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice. Raised levels of p-LIMK1/2 were also reflected in the IHC assessment. LIMK1 levels were also raised in Nisch edsn/edsn Itga5 tm1Hyn/+ mice. We were unable to assess directly levels of p-LIMK1 due to the ineffectiveness of available antibodies. While RAC1 protein levels were not affected in Nisch edsn/edsn mice, they were significantly raised in the double mutant, Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice which was mirrored in the IHC assessment. We also found by both IHC and western analysis that NF-κB levels were raised in both Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice compared to wild-type. Moreover, there is evidence from IHC that the effect of the two mutations are additive and that the double mutant shows a significantly higher level of NF-κB expression compared to Nisch edsn/edsn mice. While the protein analysis shows a similar trend the differences between Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice are not significant.
Overall, our analysis indicates that the edison mutation leads to activation of PAK1 and PAK1-independent RAC1 pathways with increased levels of NF-κB and p-LIMK1/2 each of which may lead to inflammatory and vascular permeability effects. In addition, a combination of mutations in NISCH and ITGA5 can lead to an exacerbation of raised protein levels which may underlie the more severe phenotype seen in the double mutant, Itga5 tm1Hyn/+ ; Nisch edsn/edsn . This may reflect the role of ITGA5 in enhancing binding of NISCH to PAK1, and provides us with a model of the mechanism underlying the Nisch edsn/edsn and Itga5 tm1Hyn/+ ; Nisch edsn/edsn phenotypes (Fig 8). We surmise that impairing function of NISCH leads to derepression of RAC1 pathways, while reducing levels of ITGA5 in combination with impaired NISCH leads to further derepression and activation of downstream pathways manifested in the more severe phenotype. However, we did not observe significantly raised levels of RAC1 in the single mutant. Nevertheless, we found raised levels of NF-κB that may reflect activation of PAK1 independent pathways.
ITGA5 is known to be involved with the activation of SRC protein tyrosine kinase, as well as FAK, both of which are involved with mediating VEGF induced vascular leak [32]. ITGA5 can mediate its effects by phosphorylation of FAK and binding of activated FAK to SRC leading to conformational SRC activation [33]. Alternatively, for example in the context of neuroblastoma cell motility, FAK is required for integrin α5β1-mediated SRC phosphorylation [34]. We thus investigated FAK and SRC levels in the mutant mice, focusing on activated FAK and inactive SRC. We found total levels of the full length isoform of the FAK protein raised in Nisch edsn/edsn mice. However, in Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice, full length FAK protein was not significantly different to wild-type. Importantly, we detected raised levels of activated FAK along with reduced levels of inactive SRC in Nisch edsn/edsn mice. These changes may contribute to angiogenesis and vascular permeability through upregulation of VEGF signalling pathways.
The genes mutated in previously characterised OM models (Junbo, Jeff and Tgif1), have been reported to regulate the TGF-β signalling pathway [11,12,35]. There is considerable cross-talk between TGF-β signalling and hypoxia pathways and mutations that perturb the TGF-β pathway might be expected to perturb hypoxia responses with downstream consequences on VEGFA and VEGF signalling, as is observed in the mutants studied [14]. We discuss above the raised levels of HIF-1a and VEGFA found in the edison mutant. We thus assessed TGF-β signalling (in edison), and found that p-SMAD2 levels are significantly raised in Itga5 tm1Hyn/+ ; Nisch edsn/edsn mice, but not in Nisch edsn/edsn . The lack of raised p-SMAD2 levels in the Nisch edsn/edsn mouse suggests that activation of TGF-β signalling is not a primary event underlying the development of chronic OM in the edison mouse. Rather, the upregulation of p-SMAD2 in the double mutant may reflect the very severe inflammatory state of the middle ear leading to activation of TGF-β pathways.
In summary (see Fig 8), we conclude that mutations in Nischarin impact upon PAK-dependent and PAK-independent RAC pathways with downstream signalling effects on LIMK1 and NF-κB. Moreover, the interplay between Nischarin and ITGA5 likely underlies impacts on FAK and SRC signalling which may lead to VEGF mediated vascular leak. The combined effects on LIMK1, NF-κB and FAK signalling can account for the observed inflammatory changes in the edison middle ear, which along with vascular leak and middle ear exudate, produce a chronic OM. Together, the pathways changes we describe provide a mechanism underlying the observed phenotypic changes in edison. Moreover, these studies further enhance our knowledge of the relevant genetic pathways that contribute to middle ear inflammatory disease, as well as a panel of new genes that are candidates for genetic susceptibility to chronic OM in the human population.

Ethics statement
Mice were bred and maintained by Mary Lyon Centre, MRC Harwell and were housed in specific-pathogen free conditions. All animal experimentation was approved by the Animal

Phenotype-driven ENU-mutagenesis screen
The founder mouse carrying the edison mutation was generated in a large-scale phenotypedriven ENU mutagenesis program at MRC Harwell [20]. Briefly, Male C57BL/6J mice were mutagenized and mated to C3H.Pde6b+ females (a C3H stock that does not carry the retinal degeneration allele Pde6b rd ). G 3 offspring were screened for a variety of abnormalities, including deafness and vestibular dysfunction. The edison founder was identified due to lack of a Preyer reflex when presented with a calibrated 20 kHz, 90 dB SPL tone burst via a clickbox test.

Mapping
The founder edison mouse was maintained by repeated outcrossing to C3H/HeH and intercrossing to produce homozygous mutant progeny, identified by the lack of a Preyer reflex. For linkage analysis genomic DNA from 13 affected mice were screened with 63 strain specific SNP markers spaced equidistantly across the genome using the Pyrosequencing SNP genotyping system (QIAGEN). Additional SNP markers were used within linked regions to further fine map the causal mutation. Markers and primer sequences are available on request from the authors.

Whole-genome sequencing
Genomic DNA from a single affected edison mouse was sent for next-generation sequencing (High-Throughput Genomics, WTCHG). Identification, analysis and dissemination of sequence variant information identified through whole-genome sequencing were achieved with the tools provided by the MRC Harwell Biocomputing custom sequence analysis pipeline. Sequence reads were mapped to the NCBI37/mm9 assembly of the reference mouse genome. Within the critical interval, the mean read depth was 8.17 and the read coverage was 99.64%. Sequence variants were identified and were subsequently categorised into those which occurred within exons and splice donor/acceptor sites, and were not known or strain-specific variants. This led to the identification of a T to C base substitution within exon 14 of Nisch, which is predicted to cause a Leu972Pro missense change in NISCH protein. This sequence variant was validated using Sanger sequencing (Source BioScience, UK). Affected and unaffected mice were genotyped using the LightScanner SNP genotyping system (Idaho Technology Inc., USA) to confirm the association of genotype to phenotype. In all cases the genotype correlated with the phenotype.
Gene-driven identification of an additional Nisch allele DNA from the MRC Harwell ENU-DNA sperm archive (http://www.har.mrc.ac.uk/services/ archiving-distribution/enu-dna-archive) was screened with the LightScanner platform (Idaho Technology Inc., USA). Briefly, male C57BL/6J mice were treated with ENU and crossed to C3H/HeH females. F 1 progeny (C3H/HeH.C57BL/6J) were rederived and male F 1 animals had sperm and DNA samples taken for archiving. Ten exons of Nisch were screened in DNA from 10,000 F 1 ENU mutagenised animals and potential mutations confirmed with Sanger sequencing (Source BioScience, UK).

Genetic background
All edison mice used for phenotyping were congenic on a C3H/HeH background. They were backcrossed for at least ten generations. The Nisch V33A strain was rederived by in vitro fertilisation of C57BL/6J oocytes with F 1 sperm from the MRC Harwell ENU-DNA sperm archive and maintained on a mixed C3H/HeH and C57BL/6J genetic background. Nisch V33A/+ mice were intercrossed for phenotypic analysis and crossed to congenic Nisch edsn/+ mice for complementation testing. Cryopreserved Itga5 tm1Hyn mutant sperm was imported from the Jackson Laboratory (Stock No. 002274) [48] and the colony was rederived using in vitro fertilisation of C57BL/6J oocytes. Itga5 tm1Hyn mice had been backcrossed 4-5 generations onto a C3H/HeH background, before crossing to congenic Nisch edsn mice for phenotypic analysis.

Genotyping
Genotyping for edison mice was performed using an allelic discrimination assay with primers

Skull morphology and radiography
Analysis of skulls from 20 wk mice was performed using a Faxitron Mx-20 DC-4 specimen Xray System. ImageJ software was used to measure the skull length, nasal bone length, frontal bone length, parietal bone length and skull width. Allometric comparisons were performed against skull length with at least 12 mice of each genotype.

Auditory brainstem response
Mice were anesthetized and hearing thresholds determined using ABR, as previously described [7]. The ABR threshold was measured for each ear. Click-evoked hearing assessments for edison mice were conducted at 3,

Histology
Mouse 3, 4, 6, 8, 12, 16 and 20 wk heads from Nisch +/+ , Nisch edsn/+ and Nisch edsn/edsn mice were fixed for 48 hours in 10% neutral buffered formaldehyde, decalcified in D.F.B decalcifying agent (Kristensen; Pioneer Research Chemicals) for 72 hours and embedded in paraffin following routine procedures. Perinatal heads were processed in the same manner without the decalcification steps. Four-micrometre-thick sections were obtained, de-paraffinized in xylene substitute and rehydrated via a graded ethanol. For morphological observations, sections were stained with haemotoxylin and eosin (H&E). The histological sections were used to investigate the middle ear inflammation of the mice. Evaluation of mean mucosal thickness was by blinded assessment of a standard 1000 μm length of ME mucosa (avoiding the cochlea and the region close to the Eustachian tube), the mucosal thickness was averaged from five measurements.
To study the lung morphology of edison mice, H&E stained sections from adult and perinatal lungs were viewed using a Zeiss Axiostar Plus bright-field microscope and analysed using cell B imaging software (Olympus). The data was analysed as previously described [49].

Scanning electron microscopy
To study the ultra-structure of the organ of Corti we dissected the inner ears from five 20 wk Nisch +/+ and Nisch edsn/edsn mice and prepared the samples as previously described [50]. Inner ears were imaged using a JEOL 6010 LV scanning electron microscope under high vacuum conditions.

Blood and bulla fluid collection
Blood and bulla fluids were collected, as previously described [14], for analysis using Real-time quantitative PCR.
antibodies were from the Vectastain Elite ABC kit and they were used according to the manufacturer's instructions. DAB+ chromogen system (DAKO K3468) was used to develop the specific signals. The slides were counterstained with haematoxylin.

Data analysis
All data are given as unadjusted mean +/-SEM (standard error of the mean) unless stated otherwise. Data were analysed to establish normal distribution. Where data was normally distributed an ANOVA or Student's t-test were conducted. If data was not normally distributed the non-parametric equivalents of these tests were used (Kruskal-Wallis One Way Analysis of Variance on Ranks or Mann-Whitney Rank Sum Test) to establish if data were significant. The Holm-Sidak method (ANOVA) or Dunn's method (Ranks) was used for multiple comparisons versus a control group. Results with values of P < 0.05 were considered statistically significant. SigmaPlot 11.0 software was used to perform all statistical analysis.

S1 Fig. Nisch edsn/edsn craniofacial defect has no effect on Eustachian tube morphology. (A)
Characteristic image of a male Nisch edsn/edsn mutant and a wild-type littermate. (B) Weight data for male edison mice over a 20 wk longitudinal time course shows Nisch edsn/edsn mice are significantly smaller than both wild-type and Nisch edsn/+ littermates throughout the time course. In addition, Nisch edsn/+ mice are also smaller than those wild-type for the allele. Ã P < 0.05; ÃÃ P < 0.01; ÃÃÃ P < 0.001. Nisch +/+ , n = 9; Nisch edsn/+ , n = 20; Nisch edsn/edsn , n = 18 (C) Dorsoventral view of a 20 wk wild-type mouse skull showing the measurements used to analyse skull morphology in this study. SL, skull length; SW, skull width; NB, nasal bone; FB, frontal bone; PB, parietal bone. (D) Allometric comparisons against skull length show abnormal growth in Nisch edsn/edsn skulls at the nasal bone and frontal bone. ns P > 0.05; Ã P < 0.05; ÃÃÃ P < 0.001. Nisch +/+ , n = 13; Nisch edsn/+ , n = 18; Nisch edsn/edsn , n = 12. (E) Dissected and stained skulls of Nisch +/+ , Nisch edsn/+ and Nisch edsn/edsn 20 wk mice. Eustachian tube (ET) angle measurements are indicated by dashed lines. R, right ear; L, left ear. (F) Mean angle between the midline of the skull and the bony part of the left and the right ET. ns P > 0.05. Nisch +/+ , n = 6; Nisch edsn/+ , n = 6; Nisch edsn/edsn , n = 6. Error bars indicate standard error of mean. Statistics were conducted using one-way ANOVA's and Holm-Sidak's multiple comparison tests for post-hoc analysis. (TIF) To quantify the results from the staining, airway epithelial cells in wild-types and mutants were counted in three different regions from four lungs for each genotype. The results presented from the western blots are from four independent experiments for all the antibodies except for p-LIMK1/2, where three independent experiments were used to present the data. Scale bar = 100 μm. Ã P < 0.05. Error bars indicate standard error of mean. Immunohistochemistry data was analysed by oneway ANOVAs and Holm-Sidak's multiple comparison procedures for post-hoc testing. For western blot analysis a Student's t-test was performed. (TIF)